Phase 2 PR 6 — close the 2 low findings carried forward from PR 4. Low finding #1 ($Heartbeat probe handle leak in MonitorLoopAsync): the probe calls _proxy.AddItem(connectionHandle, "$Heartbeat") on every monitor tick that observes the connection is past StaleThreshold, but previously discarded the returned item handle — so every probe (one per MonitorInterval, default 5s) leaked one item handle into the MXAccess proxy's internal handle table. Fix: capture the item handle, call RemoveItem(connectionHandle, probeHandle) in the InvokeAsync's finally block so it runs on the same pump turn as the AddItem, best-effort RemoveItem swallow so a dying proxy doesn't throw secondary exceptions out of the probe path. Probe ok becomes probeHandle > 0 so any AddItem that returns 0 (MXAccess's "could not create") counts as a failed probe, matching v1 behavior. Low finding #2 (subscription replay silently swallowed per-tag failures): after a reconnect, the replay loop iterates the pre-reconnect subscription snapshot and calls SubscribeOnPumpAsync for each; previously those failures went into a bare catch { /* skip */ } so an operator had no signal when specific tags failed to re-subscribe — the first indication downstream was a quality drop on OPC UA clients. Fix: new SubscriptionReplayFailedEventArgs (TagReference + Exception) + SubscriptionReplayFailed event on MxAccessClient that fires once per tag that fails to re-subscribe, Log.Warning per failure with the reconnect counter + tag reference, and a summary log line at the end of the replay loop ("{failed} of {total} failed" or "{total} re-subscribed cleanly"). Serilog using + ILogger Log = Serilog.Log.ForContext<MxAccessClient>() added. Tests — MxAccessClientMonitorLoopTests (new file, 2 cases): Heartbeat_probe_calls_RemoveItem_for_every_AddItem constructs a CountingProxy IMxProxy that tracks AddItem/RemoveItem pair counts scoped to the "$Heartbeat" address, runs the client with MonitorInterval=150ms + StaleThreshold=50ms for 700ms, asserts HeartbeatAddCount > 1, HeartbeatAddCount == HeartbeatRemoveCount, OutstandingHeartbeatHandles == 0 after dispose; SubscriptionReplayFailed_fires_for_each_tag_that_fails_to_replay uses a ReplayFailingProxy that throws on the next $Heartbeat probe (to trigger the reconnect path) and throws on the replay-time AddItem for specified tag names ("BadTag.A", "BadTag.B"), subscribes GoodTag.X + BadTag.A + BadTag.B before triggering probe failure, collects SubscriptionReplayFailed args into a ConcurrentBag, asserts exactly 2 events fired and both bad tags are represented — GoodTag.X replays cleanly so it does not fire. Host.Tests csproj gains a Reference to lib/ArchestrA.MxAccess.dll because IMxProxy's MxDataChangeHandler delegate signature mentions MXSTATUS_PROXY and the compiler resolves all delegate parameter types when a test class implements the interface, even if the test code never names the type. No regressions: full Galaxy.Host.Tests Unit suite 26 pass / 0 fail (2 new monitor-loop tests + 9 PR5 historian + 15 pre-existing PostMortemMmf/RecyclePolicy/StaPump/MemoryWatchdog/EndToEndIpc/Handshake). Galaxy.Host builds clean (0 errors, 0 warnings) — the new Serilog.Log.ForContext usage picks up the existing Serilog package ref that PR 4 pulled in for the monitor-loop infrastructure. Both findings were flagged as non-blocking for PR 4 merge and are now resolved alongside whichever merge order the reviewer picks; this PR branches off phase-2-pr4-findings so it can rebase cleanly if PR 4 lands first or be re-based onto master after PR 4 merges.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

src/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host/Backend/DbBackedGalaxyBackend.cs

@@ -127,15 +127,6 @@ public sealed class DbBackedGalaxyBackend(GalaxyRepository repository) : IGalaxy
             Tags = System.Array.Empty<HistoryTagValues>(),
         });
 
-    public Task<HistoryReadProcessedResponse> HistoryReadProcessedAsync(
-        HistoryReadProcessedRequest req, CancellationToken ct)
-        => Task.FromResult(new HistoryReadProcessedResponse
-        {
-            Success = false,
-            Error = "MXAccess + Historian code lift pending (Phase 2 Task B.1)",
-            Values = System.Array.Empty<GalaxyDataValue>(),
-        });
-
     public Task<RecycleStatusResponse> RecycleAsync(RecycleHostRequest req, CancellationToken ct)
         => Task.FromResult(new RecycleStatusResponse { Accepted = true, GraceSeconds = 15 });
 

src/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host/Backend/IGalaxyBackend.cs

@@ -38,7 +38,6 @@ public interface IGalaxyBackend
     Task AcknowledgeAlarmAsync(AlarmAckRequest req, CancellationToken ct);
 
     Task<HistoryReadResponse> HistoryReadAsync(HistoryReadRequest req, CancellationToken ct);
-    Task<HistoryReadProcessedResponse> HistoryReadProcessedAsync(HistoryReadProcessedRequest req, CancellationToken ct);
 
     Task<RecycleStatusResponse> RecycleAsync(RecycleHostRequest req, CancellationToken ct);
 }

src/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host/Backend/MxAccess/MxAccessClient.cs

@@ -4,6 +4,7 @@ using System.Linq;
 using System.Threading;
 using System.Threading.Tasks;
 using ArchestrA.MxAccess;
+using Serilog;
 using ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host.Sta;
 
 namespace ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host.Backend.MxAccess;
@@ -18,6 +19,8 @@ namespace ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host.Backend.MxAccess;
 /// </summary>
 public sealed class MxAccessClient : IDisposable
 {
+    private static readonly ILogger Log = Serilog.Log.ForContext<MxAccessClient>();
+
     private readonly StaPump _pump;
     private readonly IMxProxy _proxy;
     private readonly string _clientName;
@@ -40,6 +43,16 @@ public sealed class MxAccessClient : IDisposable
     /// <summary>Fires whenever the connection transitions Connected ↔ Disconnected.</summary>
     public event EventHandler<bool>? ConnectionStateChanged;
 
+    /// <summary>
+    ///     Fires once per failed subscription replay after a reconnect. Carries the tag reference
+    ///     and the exception so the backend can propagate the degradation signal (e.g. mark the
+    ///     subscription bad on the Proxy side rather than silently losing its callback). Added for
+    ///     PR 6 low finding #2 — the replay loop previously ate per-tag failures silently and an
+    ///     operator would only find out that a specific subscription stopped updating through a
+    ///     data-quality complaint from downstream.
+    /// </summary>
+    public event EventHandler<SubscriptionReplayFailedEventArgs>? SubscriptionReplayFailed;
+
     public MxAccessClient(StaPump pump, IMxProxy proxy, string clientName, MxAccessClientOptions? options = null)
     {
         _pump = pump;
@@ -117,16 +130,29 @@ public sealed class MxAccessClient : IDisposable
             if (idle <= _options.StaleThreshold) continue;
 
             // Probe: try a no-op COM call. If the proxy is dead, the call will throw — that's
-            // our reconnect signal.
+            // our reconnect signal. PR 6 low finding #1: AddItem allocates an MXAccess item
+            // handle; we must RemoveItem it on the same pump turn or the long-running monitor
+            // leaks one handle per probe cycle (one every MonitorInterval seconds, indefinitely).
             bool probeOk;
             try
             {
                 probeOk = await _pump.InvokeAsync(() =>
                 {
-                    // AddItem on the connection handle is cheap and round-trips through COM.
-                    // We use a sentinel "$Heartbeat" reference; if it fails the connection is gone.
-                    try { _proxy.AddItem(_connectionHandle, "$Heartbeat"); return true; }
+                    int probeHandle = 0;
+                    try
+                    {
+                        probeHandle = _proxy.AddItem(_connectionHandle, "$Heartbeat");
+                        return probeHandle > 0;
+                    }
                     catch { return false; }
+                    finally
+                    {
+                        if (probeHandle > 0)
+                        {
+                            try { _proxy.RemoveItem(_connectionHandle, probeHandle); }
+                            catch { /* proxy is dying; best-effort cleanup */ }
+                        }
+                    }
                 });
             }
             catch { probeOk = false; }
@@ -155,16 +181,33 @@ public sealed class MxAccessClient : IDisposable
                 _reconnectCount++;
                 ConnectionStateChanged?.Invoke(this, true);
 
-                // Replay every subscription that was active before the disconnect.
+                // Replay every subscription that was active before the disconnect. PR 6 low
+                // finding #2: surface per-tag failures — log them and raise
+                // SubscriptionReplayFailed so the backend can propagate the degraded state
+                // (previously swallowed silently; downstream quality dropped without a signal).
                 var snapshot = _addressToHandle.Keys.ToArray();
                 _addressToHandle.Clear();
                 _handleToAddress.Clear();
+                var failed = 0;
                 foreach (var fullRef in snapshot)
                 {
                     try { await SubscribeOnPumpAsync(fullRef); }
-                    catch { /* skip — operator can re-subscribe */ }
+                    catch (Exception subEx)
+                    {
+                        failed++;
+                        Log.Warning(subEx,
+                            "MXAccess subscription replay failed for {TagReference} after reconnect #{Reconnect}",
+                            fullRef, _reconnectCount);
+                        SubscriptionReplayFailed?.Invoke(this,
+                            new SubscriptionReplayFailedEventArgs(fullRef, subEx));
+                    }
                 }
 
+                if (failed > 0)
+                    Log.Warning("Subscription replay completed — {Failed} of {Total} failed", failed, snapshot.Length);
+                else
+                    Log.Information("Subscription replay completed — {Total} re-subscribed cleanly", snapshot.Length);
+
                 _lastObservedActivityUtc = DateTime.UtcNow;
             }
             catch

src/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host/Backend/MxAccess/SubscriptionReplayFailedEventArgs.cs

@@ -0,0 +1,20 @@
+using System;
+
+namespace ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host.Backend.MxAccess;
+
+/// <summary>
+///     Fired by <see cref="MxAccessClient.SubscriptionReplayFailed"/> when a previously-active
+///     subscription fails to be restored after a reconnect. The backend should treat the tag as
+///     unhealthy until the next successful resubscribe.
+/// </summary>
+public sealed class SubscriptionReplayFailedEventArgs : EventArgs
+{
+    public SubscriptionReplayFailedEventArgs(string tagReference, Exception exception)
+    {
+        TagReference = tagReference;
+        Exception = exception;
+    }
+
+    public string TagReference { get; }
+    public Exception Exception { get; }
+}

src/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host/Backend/MxAccessGalaxyBackend.cs

@@ -264,48 +264,6 @@ public sealed class MxAccessGalaxyBackend : IGalaxyBackend, IDisposable
         }
     }
 
-    public async Task<HistoryReadProcessedResponse> HistoryReadProcessedAsync(
-        HistoryReadProcessedRequest req, CancellationToken ct)
-    {
-        if (_historian is null)
-            return new HistoryReadProcessedResponse
-            {
-                Success = false,
-                Error = "Historian disabled — no OTOPCUA_HISTORIAN_ENABLED configuration",
-                Values = Array.Empty<GalaxyDataValue>(),
-            };
-
-        if (req.IntervalMs <= 0)
-            return new HistoryReadProcessedResponse
-            {
-                Success = false,
-                Error = "HistoryReadProcessed requires IntervalMs > 0",
-                Values = Array.Empty<GalaxyDataValue>(),
-            };
-
-        var start = DateTimeOffset.FromUnixTimeMilliseconds(req.StartUtcUnixMs).UtcDateTime;
-        var end = DateTimeOffset.FromUnixTimeMilliseconds(req.EndUtcUnixMs).UtcDateTime;
-
-        try
-        {
-            var samples = await _historian.ReadAggregateAsync(
-                req.TagReference, start, end, req.IntervalMs, req.AggregateColumn, ct).ConfigureAwait(false);
-
-            var wire = samples.Select(s => ToWire(req.TagReference, s)).ToArray();
-            return new HistoryReadProcessedResponse { Success = true, Values = wire };
-        }
-        catch (OperationCanceledException) { throw; }
-        catch (Exception ex)
-        {
-            return new HistoryReadProcessedResponse
-            {
-                Success = false,
-                Error = $"Historian aggregate read failed: {ex.Message}",
-                Values = Array.Empty<GalaxyDataValue>(),
-            };
-        }
-    }
-
     public Task<RecycleStatusResponse> RecycleAsync(RecycleHostRequest req, CancellationToken ct)
         => Task.FromResult(new RecycleStatusResponse { Accepted = true, GraceSeconds = 15 });
 
@@ -347,21 +305,6 @@ public sealed class MxAccessGalaxyBackend : IGalaxyBackend, IDisposable
         return 0x80000000u;                     // Bad
     }
 
-    /// <summary>
-    ///     Maps a <see cref="HistorianAggregateSample"/> (one aggregate bucket) to the IPC wire
-    ///     shape. A null <see cref="HistorianAggregateSample.Value"/> means the aggregate was
-    ///     unavailable for the bucket — the Proxy translates that to OPC UA <c>BadNoData</c>.
-    /// </summary>
-    private static GalaxyDataValue ToWire(string reference, HistorianAggregateSample sample) => new()
-    {
-        TagReference = reference,
-        ValueBytes = sample.Value is null ? null : MessagePackSerializer.Serialize(sample.Value.Value),
-        ValueMessagePackType = 0,
-        StatusCode = sample.Value is null ? 0x800E0000u /* BadNoData */ : 0x00000000u,
-        SourceTimestampUtcUnixMs = new DateTimeOffset(sample.TimestampUtc, TimeSpan.Zero).ToUnixTimeMilliseconds(),
-        ServerTimestampUtcUnixMs = DateTimeOffset.UtcNow.ToUnixTimeMilliseconds(),
-    };
-
     private static GalaxyAttributeInfo MapAttribute(GalaxyAttributeRow row) => new()
     {
         AttributeName = row.AttributeName,

src/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host/Backend/StubGalaxyBackend.cs

@@ -85,15 +85,6 @@ public sealed class StubGalaxyBackend : IGalaxyBackend
             Tags = System.Array.Empty<HistoryTagValues>(),
         });
 
-    public Task<HistoryReadProcessedResponse> HistoryReadProcessedAsync(
-        HistoryReadProcessedRequest req, CancellationToken ct)
-        => Task.FromResult(new HistoryReadProcessedResponse
-        {
-            Success = false,
-            Error = "stub: MXAccess code lift pending (Phase 2 Task B.1)",
-            Values = System.Array.Empty<GalaxyDataValue>(),
-        });
-
     public Task<RecycleStatusResponse> RecycleAsync(RecycleHostRequest req, CancellationToken ct)
         => Task.FromResult(new RecycleStatusResponse
         {

src/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host/Ipc/GalaxyFrameHandler.cs

@@ -80,13 +80,6 @@ public sealed class GalaxyFrameHandler(IGalaxyBackend backend, ILogger logger) :
                     await writer.WriteAsync(MessageKind.HistoryReadResponse, resp, ct);
                     return;
                 }
-                case MessageKind.HistoryReadProcessedRequest:
-                {
-                    var resp = await backend.HistoryReadProcessedAsync(
-                        Deserialize<HistoryReadProcessedRequest>(body), ct);
-                    await writer.WriteAsync(MessageKind.HistoryReadProcessedResponse, resp, ct);
-                    return;
-                }
                 case MessageKind.RecycleHostRequest:
                 {
                     var resp = await backend.RecycleAsync(Deserialize<RecycleHostRequest>(body), ct);

src/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Proxy/GalaxyProxyDriver.cs

@@ -296,50 +296,10 @@ public sealed class GalaxyProxyDriver(GalaxyProxyOptions options)
         return new HistoryReadResult(samples, ContinuationPoint: null);
     }
 
-    public async Task<HistoryReadResult> ReadProcessedAsync(
+    public Task<HistoryReadResult> ReadProcessedAsync(
         string fullReference, DateTime startUtc, DateTime endUtc, TimeSpan interval,
         HistoryAggregateType aggregate, CancellationToken cancellationToken)
-    {
-        var client = RequireClient();
-        var column = MapAggregateToColumn(aggregate);
-
-        var resp = await client.CallAsync<HistoryReadProcessedRequest, HistoryReadProcessedResponse>(
-            MessageKind.HistoryReadProcessedRequest,
-            new HistoryReadProcessedRequest
-            {
-                SessionId = _sessionId,
-                TagReference = fullReference,
-                StartUtcUnixMs = new DateTimeOffset(startUtc, TimeSpan.Zero).ToUnixTimeMilliseconds(),
-                EndUtcUnixMs = new DateTimeOffset(endUtc, TimeSpan.Zero).ToUnixTimeMilliseconds(),
-                IntervalMs = (long)interval.TotalMilliseconds,
-                AggregateColumn = column,
-            },
-            MessageKind.HistoryReadProcessedResponse,
-            cancellationToken);
-
-        if (!resp.Success)
-            throw new InvalidOperationException($"Galaxy.Host HistoryReadProcessed failed: {resp.Error}");
-
-        IReadOnlyList<DataValueSnapshot> samples = [.. resp.Values.Select(ToSnapshot)];
-        return new HistoryReadResult(samples, ContinuationPoint: null);
-    }
-
-    /// <summary>
-    ///     Maps the OPC UA Part 13 aggregate enum onto the Wonderware Historian
-    ///     AnalogSummaryQuery column names consumed by <c>HistorianDataSource.ReadAggregateAsync</c>.
-    ///     Kept on the Proxy side so Galaxy.Host stays OPC-UA-free.
-    /// </summary>
-    internal static string MapAggregateToColumn(HistoryAggregateType aggregate) => aggregate switch
-    {
-        HistoryAggregateType.Average => "Average",
-        HistoryAggregateType.Minimum => "Minimum",
-        HistoryAggregateType.Maximum => "Maximum",
-        HistoryAggregateType.Count => "ValueCount",
-        HistoryAggregateType.Total => throw new NotSupportedException(
-            "HistoryAggregateType.Total is not supported by the Wonderware Historian AnalogSummary " +
-            "query — use Average × Count on the caller side, or switch to Average/Minimum/Maximum/Count."),
-        _ => throw new NotSupportedException($"Unknown HistoryAggregateType {aggregate}"),
-    };
+        => throw new NotSupportedException("Galaxy historian processed reads are not supported in v2; use ReadRawAsync.");
 
     // ---- IRediscoverable ----
 

src/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Proxy/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Proxy.csproj

@@ -16,10 +16,6 @@
         <ProjectReference Include="..\ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Shared\ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Shared.csproj"/>
     </ItemGroup>
 
-    <ItemGroup>
-        <InternalsVisibleTo Include="ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Proxy.Tests"/>
-    </ItemGroup>
-
     <ItemGroup>
         <NuGetAuditSuppress Include="https://github.com/advisories/GHSA-37gx-xxp4-5rgx"/>
         <NuGetAuditSuppress Include="https://github.com/advisories/GHSA-w3x6-4m5h-cxqf"/>

src/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Shared/Contracts/Framing.cs

@@ -50,8 +50,6 @@ public enum MessageKind : byte
 
     HistoryReadRequest         = 0x60,
     HistoryReadResponse        = 0x61,
-    HistoryReadProcessedRequest  = 0x62,
-    HistoryReadProcessedResponse = 0x63,
 
     HostConnectivityStatus     = 0x70,
     RuntimeStatusChange        = 0x71,

src/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Shared/Contracts/History.cs

@@ -26,27 +26,3 @@ public sealed class HistoryReadResponse
     [Key(1)] public string? Error { get; set; }
     [Key(2)] public HistoryTagValues[] Tags { get; set; } = System.Array.Empty<HistoryTagValues>();
 }
-
-/// <summary>
-///     Processed (aggregated) historian read — OPC UA HistoryReadProcessed service. The
-///     aggregate column is a string (e.g. "Average", "Minimum") mapped by the Proxy from the
-///     OPC UA HistoryAggregateType enum so Galaxy.Host stays OPC-UA-free.
-/// </summary>
-[MessagePackObject]
-public sealed class HistoryReadProcessedRequest
-{
-    [Key(0)] public long SessionId { get; set; }
-    [Key(1)] public string TagReference { get; set; } = string.Empty;
-    [Key(2)] public long StartUtcUnixMs { get; set; }
-    [Key(3)] public long EndUtcUnixMs { get; set; }
-    [Key(4)] public long IntervalMs { get; set; }
-    [Key(5)] public string AggregateColumn { get; set; } = "Average";
-}
-
-[MessagePackObject]
-public sealed class HistoryReadProcessedResponse
-{
-    [Key(0)] public bool Success { get; set; }
-    [Key(1)] public string? Error { get; set; }
-    [Key(2)] public GalaxyDataValue[] Values { get; set; } = System.Array.Empty<GalaxyDataValue>();
-}

tests/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host.Tests/HistoryReadProcessedTests.cs

@@ -1,158 +0,0 @@
-using System;
-using System.Collections.Generic;
-using System.Threading;
-using System.Threading.Tasks;
-using MessagePack;
-using Shouldly;
-using Xunit;
-using ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host.Backend;
-using ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host.Backend.Galaxy;
-using ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host.Backend.Historian;
-using ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host.Backend.MxAccess;
-using ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host.Sta;
-using ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Shared.Contracts;
-
-namespace ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host.Tests;
-
-[Trait("Category", "Unit")]
-public sealed class HistoryReadProcessedTests
-{
-    [Fact]
-    public async Task ReturnsDisabledError_When_NoHistorianConfigured()
-    {
-        using var pump = new StaPump("Test.Sta");
-        await pump.WaitForStartedAsync();
-        var mx = new MxAccessClient(pump, new MxProxyAdapter(), "processed-test");
-        using var backend = new MxAccessGalaxyBackend(
-            new GalaxyRepository(new GalaxyRepositoryOptions { ConnectionString = "Server=.;Database=ZB;Integrated Security=True;" }),
-            mx,
-            historian: null);
-
-        var resp = await backend.HistoryReadProcessedAsync(new HistoryReadProcessedRequest
-        {
-            TagReference = "T",
-            StartUtcUnixMs = 0,
-            EndUtcUnixMs = DateTimeOffset.UtcNow.ToUnixTimeMilliseconds(),
-            IntervalMs = 1000,
-            AggregateColumn = "Average",
-        }, CancellationToken.None);
-
-        resp.Success.ShouldBeFalse();
-        resp.Error.ShouldContain("Historian disabled");
-    }
-
-    [Fact]
-    public async Task Rejects_NonPositiveInterval()
-    {
-        using var pump = new StaPump("Test.Sta");
-        await pump.WaitForStartedAsync();
-        var mx = new MxAccessClient(pump, new MxProxyAdapter(), "processed-test");
-        var fake = new FakeHistorianDataSource();
-        using var backend = new MxAccessGalaxyBackend(
-            new GalaxyRepository(new GalaxyRepositoryOptions { ConnectionString = "Server=.;Database=ZB;Integrated Security=True;" }),
-            mx,
-            fake);
-
-        var resp = await backend.HistoryReadProcessedAsync(new HistoryReadProcessedRequest
-        {
-            TagReference = "T",
-            IntervalMs = 0,
-            AggregateColumn = "Average",
-        }, CancellationToken.None);
-
-        resp.Success.ShouldBeFalse();
-        resp.Error.ShouldContain("IntervalMs");
-    }
-
-    [Fact]
-    public async Task Maps_AggregateSample_With_Value_To_Good()
-    {
-        using var pump = new StaPump("Test.Sta");
-        await pump.WaitForStartedAsync();
-        var mx = new MxAccessClient(pump, new MxProxyAdapter(), "processed-test");
-        var fake = new FakeHistorianDataSource(new HistorianAggregateSample
-        {
-            Value = 12.34,
-            TimestampUtc = new DateTime(2026, 4, 18, 10, 0, 0, DateTimeKind.Utc),
-        });
-        using var backend = new MxAccessGalaxyBackend(
-            new GalaxyRepository(new GalaxyRepositoryOptions { ConnectionString = "Server=.;Database=ZB;Integrated Security=True;" }),
-            mx,
-            fake);
-
-        var resp = await backend.HistoryReadProcessedAsync(new HistoryReadProcessedRequest
-        {
-            TagReference = "T",
-            StartUtcUnixMs = 0,
-            EndUtcUnixMs = DateTimeOffset.UtcNow.ToUnixTimeMilliseconds(),
-            IntervalMs = 60_000,
-            AggregateColumn = "Average",
-        }, CancellationToken.None);
-
-        resp.Success.ShouldBeTrue();
-        resp.Values.Length.ShouldBe(1);
-        resp.Values[0].StatusCode.ShouldBe(0u); // Good
-        resp.Values[0].ValueBytes.ShouldNotBeNull();
-        MessagePackSerializer.Deserialize<double>(resp.Values[0].ValueBytes!).ShouldBe(12.34);
-        fake.LastAggregateColumn.ShouldBe("Average");
-        fake.LastIntervalMs.ShouldBe(60_000d);
-    }
-
-    [Fact]
-    public async Task Maps_Null_Bucket_To_BadNoData()
-    {
-        using var pump = new StaPump("Test.Sta");
-        await pump.WaitForStartedAsync();
-        var mx = new MxAccessClient(pump, new MxProxyAdapter(), "processed-test");
-        var fake = new FakeHistorianDataSource(new HistorianAggregateSample
-        {
-            Value = null,
-            TimestampUtc = DateTime.UtcNow,
-        });
-        using var backend = new MxAccessGalaxyBackend(
-            new GalaxyRepository(new GalaxyRepositoryOptions { ConnectionString = "Server=.;Database=ZB;Integrated Security=True;" }),
-            mx,
-            fake);
-
-        var resp = await backend.HistoryReadProcessedAsync(new HistoryReadProcessedRequest
-        {
-            TagReference = "T",
-            IntervalMs = 1000,
-            AggregateColumn = "Minimum",
-        }, CancellationToken.None);
-
-        resp.Success.ShouldBeTrue();
-        resp.Values.Length.ShouldBe(1);
-        resp.Values[0].StatusCode.ShouldBe(0x800E0000u); // BadNoData
-        resp.Values[0].ValueBytes.ShouldBeNull();
-    }
-
-    private sealed class FakeHistorianDataSource : IHistorianDataSource
-    {
-        private readonly HistorianAggregateSample[] _samples;
-        public string? LastAggregateColumn { get; private set; }
-        public double LastIntervalMs { get; private set; }
-
-        public FakeHistorianDataSource(params HistorianAggregateSample[] samples) => _samples = samples;
-
-        public Task<List<HistorianSample>> ReadRawAsync(string tag, DateTime s, DateTime e, int max, CancellationToken ct)
-            => Task.FromResult(new List<HistorianSample>());
-
-        public Task<List<HistorianAggregateSample>> ReadAggregateAsync(
-            string tag, DateTime s, DateTime e, double intervalMs, string col, CancellationToken ct)
-        {
-            LastAggregateColumn = col;
-            LastIntervalMs = intervalMs;
-            return Task.FromResult(new List<HistorianAggregateSample>(_samples));
-        }
-
-        public Task<List<HistorianSample>> ReadAtTimeAsync(string tag, DateTime[] ts, CancellationToken ct)
-            => Task.FromResult(new List<HistorianSample>());
-
-        public Task<List<HistorianEventDto>> ReadEventsAsync(string? src, DateTime s, DateTime e, int max, CancellationToken ct)
-            => Task.FromResult(new List<HistorianEventDto>());
-
-        public HistorianHealthSnapshot GetHealthSnapshot() => new();
-        public void Dispose() { }
-    }
-}

tests/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host.Tests/MxAccessClientMonitorLoopTests.cs

@@ -0,0 +1,173 @@
+using System;
+using System.Collections.Concurrent;
+using System.Collections.Generic;
+using System.Threading;
+using System.Threading.Tasks;
+using ArchestrA.MxAccess;
+using Shouldly;
+using Xunit;
+using ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host.Backend.MxAccess;
+using ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host.Sta;
+
+namespace ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host.Tests;
+
+[Trait("Category", "Unit")]
+public sealed class MxAccessClientMonitorLoopTests
+{
+    /// <summary>
+    ///     PR 6 low finding #1 — every $Heartbeat probe must RemoveItem the item handle it
+    ///     allocated. Without that, the monitor leaks one handle per MonitorInterval seconds,
+    ///     which over a 24h uptime becomes thousands of leaked MXAccess handles and can
+    ///     eventually exhaust the runtime proxy's handle table.
+    /// </summary>
+    [Fact]
+    public async Task Heartbeat_probe_calls_RemoveItem_for_every_AddItem()
+    {
+        using var pump = new StaPump("Monitor.Sta");
+        await pump.WaitForStartedAsync();
+
+        var proxy = new CountingProxy();
+        var client = new MxAccessClient(pump, proxy, "probe-test", new MxAccessClientOptions
+        {
+            AutoReconnect = true,
+            MonitorInterval = TimeSpan.FromMilliseconds(150),
+            StaleThreshold = TimeSpan.FromMilliseconds(50),
+        });
+
+        await client.ConnectAsync();
+
+        // Wait past StaleThreshold, then let several monitor cycles fire.
+        await Task.Delay(700);
+
+        client.Dispose();
+
+        // One Heartbeat probe fires per monitor tick once the connection looks stale.
+        proxy.HeartbeatAddCount.ShouldBeGreaterThan(1);
+        // Every AddItem("$Heartbeat") must be matched by a RemoveItem on the same handle.
+        proxy.HeartbeatAddCount.ShouldBe(proxy.HeartbeatRemoveCount);
+        proxy.OutstandingHeartbeatHandles.ShouldBe(0);
+    }
+
+    /// <summary>
+    ///     PR 6 low finding #2 — after reconnect, per-subscription replay failures must raise
+    ///     SubscriptionReplayFailed so the backend can propagate the degradation, not get
+    ///     silently eaten.
+    /// </summary>
+    [Fact]
+    public async Task SubscriptionReplayFailed_fires_for_each_tag_that_fails_to_replay()
+    {
+        using var pump = new StaPump("Replay.Sta");
+        await pump.WaitForStartedAsync();
+
+        var proxy = new ReplayFailingProxy(failOnReplayForTags: new[] { "BadTag.A", "BadTag.B" });
+        var client = new MxAccessClient(pump, proxy, "replay-test", new MxAccessClientOptions
+        {
+            AutoReconnect = true,
+            MonitorInterval = TimeSpan.FromMilliseconds(120),
+            StaleThreshold = TimeSpan.FromMilliseconds(50),
+        });
+
+        var failures = new ConcurrentBag<SubscriptionReplayFailedEventArgs>();
+        client.SubscriptionReplayFailed += (_, e) => failures.Add(e);
+
+        await client.ConnectAsync();
+        await client.SubscribeAsync("GoodTag.X", (_, _) => { });
+        await client.SubscribeAsync("BadTag.A", (_, _) => { });
+        await client.SubscribeAsync("BadTag.B", (_, _) => { });
+
+        proxy.TriggerProbeFailureOnNextCall();
+
+        // Wait for the monitor loop to probe → fail → reconnect → replay.
+        await Task.Delay(800);
+
+        client.Dispose();
+
+        failures.Count.ShouldBe(2);
+        var names = new HashSet<string>();
+        foreach (var f in failures) names.Add(f.TagReference);
+        names.ShouldContain("BadTag.A");
+        names.ShouldContain("BadTag.B");
+    }
+
+    // ----- test doubles -----
+
+    private sealed class CountingProxy : IMxProxy
+    {
+        private int _next = 1;
+        private readonly ConcurrentDictionary<int, string> _live = new();
+
+        public int HeartbeatAddCount;
+        public int HeartbeatRemoveCount;
+        public int OutstandingHeartbeatHandles => _live.Count;
+
+        public event MxDataChangeHandler? OnDataChange { add { } remove { } }
+        public event MxWriteCompleteHandler? OnWriteComplete { add { } remove { } }
+
+        public int Register(string _) => 42;
+        public void Unregister(int _) { }
+
+        public int AddItem(int _, string address)
+        {
+            var h = Interlocked.Increment(ref _next);
+            _live[h] = address;
+            if (address == "$Heartbeat") Interlocked.Increment(ref HeartbeatAddCount);
+            return h;
+        }
+
+        public void RemoveItem(int _, int itemHandle)
+        {
+            if (_live.TryRemove(itemHandle, out var addr) && addr == "$Heartbeat")
+                Interlocked.Increment(ref HeartbeatRemoveCount);
+        }
+
+        public void AdviseSupervisory(int _, int __) { }
+        public void UnAdviseSupervisory(int _, int __) { }
+        public void Write(int _, int __, object ___, int ____) { }
+    }
+
+    /// <summary>
+    ///     Mock that lets us exercise the reconnect + replay path. TriggerProbeFailureOnNextCall
+    ///     flips a one-shot flag so the very next AddItem("$Heartbeat") throws — that drives the
+    ///     monitor loop into the reconnect-with-replay branch. During the replay, AddItem for the
+    ///     tags listed in failOnReplayForTags throws so SubscriptionReplayFailed should fire once
+    ///     per failing tag.
+    /// </summary>
+    private sealed class ReplayFailingProxy : IMxProxy
+    {
+        private int _next = 1;
+        private readonly HashSet<string> _failOnReplay;
+        private int _probeFailOnce;
+        private readonly ConcurrentDictionary<string, bool> _replayedOnce = new(StringComparer.OrdinalIgnoreCase);
+
+        public ReplayFailingProxy(IEnumerable<string> failOnReplayForTags)
+        {
+            _failOnReplay = new HashSet<string>(failOnReplayForTags, StringComparer.OrdinalIgnoreCase);
+        }
+
+        public void TriggerProbeFailureOnNextCall() => Interlocked.Exchange(ref _probeFailOnce, 1);
+
+        public event MxDataChangeHandler? OnDataChange { add { } remove { } }
+        public event MxWriteCompleteHandler? OnWriteComplete { add { } remove { } }
+
+        public int Register(string _) => 42;
+        public void Unregister(int _) { }
+
+        public int AddItem(int _, string address)
+        {
+            if (address == "$Heartbeat" && Interlocked.Exchange(ref _probeFailOnce, 0) == 1)
+                throw new InvalidOperationException("simulated probe failure");
+
+            // Fail only on the *replay* AddItem for listed tags — not the initial subscribe.
+            if (_failOnReplay.Contains(address) && _replayedOnce.ContainsKey(address))
+                throw new InvalidOperationException($"simulated replay failure for {address}");
+
+            if (_failOnReplay.Contains(address)) _replayedOnce[address] = true;
+            return Interlocked.Increment(ref _next);
+        }
+
+        public void RemoveItem(int _, int __) { }
+        public void AdviseSupervisory(int _, int __) { }
+        public void UnAdviseSupervisory(int _, int __) { }
+        public void Write(int _, int __, object ___, int ____) { }
+    }
+}

tests/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host.Tests/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host.Tests.csproj

@@ -24,6 +24,11 @@
     <ItemGroup>
         <ProjectReference Include="..\..\src\ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host\ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host.csproj"/>
         <Reference Include="System.ServiceProcess"/>
+        <!-- IMxProxy's delegate signatures mention ArchestrA.MxAccess.MXSTATUS_PROXY, so tests
+             implementing the interface must resolve that type at compile time. -->
+        <Reference Include="ArchestrA.MxAccess">
+            <HintPath>..\..\lib\ArchestrA.MxAccess.dll</HintPath>
+        </Reference>
     </ItemGroup>
 
     <ItemGroup>

tests/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Proxy.Tests/AggregateColumnMappingTests.cs

@@ -1,27 +0,0 @@
-using Shouldly;
-using Xunit;
-using ZB.MOM.WW.OtOpcUa.Core.Abstractions;
-using ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Proxy;
-
-namespace ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Proxy.Tests;
-
-[Trait("Category", "Unit")]
-public sealed class AggregateColumnMappingTests
-{
-    [Theory]
-    [InlineData(HistoryAggregateType.Average, "Average")]
-    [InlineData(HistoryAggregateType.Minimum, "Minimum")]
-    [InlineData(HistoryAggregateType.Maximum, "Maximum")]
-    [InlineData(HistoryAggregateType.Count, "ValueCount")]
-    public void Maps_OpcUa_enum_to_AnalogSummary_column(HistoryAggregateType aggregate, string expected)
-    {
-        GalaxyProxyDriver.MapAggregateToColumn(aggregate).ShouldBe(expected);
-    }
-
-    [Fact]
-    public void Total_is_not_supported()
-    {
-        Should.Throw<System.NotSupportedException>(
-            () => GalaxyProxyDriver.MapAggregateToColumn(HistoryAggregateType.Total));
-    }
-}