AlarmSurfaceInvoker — wraps IAlarmSource.Subscribe/Unsubscribe/Acknowledge through CapabilityInvoker with multi-host fan-out. Closes alarm-surface slice of task #161 (Phase 6.1 Stream A); the Roslyn invoker-coverage analyzer is split into new task #200 because a DiagnosticAnalyzer project is genuinely its own scaffolding PR (Microsoft.CodeAnalysis.CSharp.Workspaces dep, netstandard2.0 target, Microsoft.CodeAnalysis.Testing harness, ProjectReference OutputItemType=Analyzer wiring, and four corner-case rules I want tests for before shipping). Ship this PR as the runtime guardrail + callable API; the analyzer lands next as the compile-time guardrail. New AlarmSurfaceInvoker class in Core.Resilience. Three methods mirror IAlarmSource's three mutating surfaces: SubscribeAsync (fan-out: group sourceNodeIds by IPerCallHostResolver.ResolveHost, one CapabilityInvoker.ExecuteAsync per host with DriverCapability.AlarmSubscribe so AlarmSubscribe's retry policy kicks in + returns one IAlarmSubscriptionHandle per host); UnsubscribeAsync (single-host, defaultHost); AcknowledgeAsync (fan-out: group AlarmAcknowledgeRequests by resolver-mapped host, run each host's batch through DriverCapability.AlarmAcknowledge which does NOT retry per decision #143 — alarm-ack is a write-shaped op that's not idempotent at the plant-floor level). Drivers without IPerCallHostResolver (Galaxy single MXAccess endpoint, OpcUaClient against one remote, etc.) fall back to defaultHost = DriverInstanceId so breaker + bulkhead keying still happens; drivers with it get one-dead-PLC-doesn't-poison-siblings isolation per decision #144. Single-host single-subscribe returns [handle] with length 1; empty sourceNodeIds fast-paths to [] without a driver call. Five new AlarmSurfaceInvokerTests covering: (a) empty list short-circuits — driver method never called; (b) single-host sub routes via default host — one driver call with full id list; (c) multi-host sub fans out to 2 distinct hosts for 3 src ids mapping to 2 plcs — one driver call per host; (d) Acknowledge does not retry on failure — call count stays at 1 even with exception; (e) Subscribe retries transient failures — call count reaches 3 with a 2-failures-then-success fake. Core.Tests resilience-builder suite 19/19 passing (was 14, +5); Core.Tests whole suite still green. Core project builds 0 errors. Task #200 captures the compile-time guardrail: Roslyn DiagnosticAnalyzer at src/ZB.MOM.WW.OtOpcUa.Analyzers that flags direct invocations of the eleven capability-interface methods inside the Server namespace when the call is NOT inside a CapabilityInvoker.ExecuteAsync/ExecuteWriteAsync/AlarmSurfaceInvoker.*Async lambda. That analyzer is the reason we keep paying the wrapping-class overhead for every new capability.

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

src/ZB.MOM.WW.OtOpcUa.Admin/Services/EquipmentImportBatchService.cs

@@ -2,6 +2,7 @@ using Microsoft.EntityFrameworkCore;
 using ZB.MOM.WW.OtOpcUa.Admin.Services;
 using ZB.MOM.WW.OtOpcUa.Configuration;
 using ZB.MOM.WW.OtOpcUa.Configuration.Entities;
+using ZB.MOM.WW.OtOpcUa.Configuration.Enums;
 
 namespace ZB.MOM.WW.OtOpcUa.Admin.Services;
 
@@ -152,14 +153,37 @@ public sealed class EquipmentImportBatchService(OtOpcUaConfigDbContext db)
 
         try
         {
-            foreach (var row in batch.Rows.Where(r => r.IsAccepted))
+            // Snapshot active reservations that overlap this batch's ZTag + SAPID set — one
+            // round-trip instead of N. Released rows (ReleasedAt IS NOT NULL) are ignored so
+            // an explicitly-released value can be reused.
+            var accepted = batch.Rows.Where(r => r.IsAccepted).ToList();
+            var zTags = accepted.Where(r => !string.IsNullOrWhiteSpace(r.ZTag))
+                .Select(r => r.ZTag).Distinct(StringComparer.OrdinalIgnoreCase).ToList();
+            var sapIds = accepted.Where(r => !string.IsNullOrWhiteSpace(r.SAPID))
+                .Select(r => r.SAPID).Distinct(StringComparer.OrdinalIgnoreCase).ToList();
+
+            var existingReservations = await db.ExternalIdReservations
+                .Where(r => r.ReleasedAt == null &&
+                    ((r.Kind == ReservationKind.ZTag && zTags.Contains(r.Value)) ||
+                     (r.Kind == ReservationKind.SAPID && sapIds.Contains(r.Value))))
+                .ToListAsync(ct).ConfigureAwait(false);
+            var resByKey = existingReservations.ToDictionary(
+                r => (r.Kind, r.Value.ToLowerInvariant()),
+                r => r);
+
+            var nowUtc = DateTime.UtcNow;
+            var firstPublishedBy = batch.CreatedBy;
+
+            foreach (var row in accepted)
             {
+                var equipmentUuid = Guid.TryParse(row.EquipmentUuid, out var u) ? u : Guid.NewGuid();
+
                 db.Equipment.Add(new Equipment
                 {
                     EquipmentRowId = Guid.NewGuid(),
                     GenerationId = generationId,
                     EquipmentId = row.EquipmentId,
-                    EquipmentUuid = Guid.TryParse(row.EquipmentUuid, out var u) ? u : Guid.NewGuid(),
+                    EquipmentUuid = equipmentUuid,
                     DriverInstanceId = driverInstanceIdForRows,
                     UnsLineId = unsLineIdForRows,
                     Name = row.Name,
@@ -176,10 +200,25 @@ public sealed class EquipmentImportBatchService(OtOpcUaConfigDbContext db)
                     ManufacturerUri = row.ManufacturerUri,
                     DeviceManualUri = row.DeviceManualUri,
                 });
+
+                MergeReservation(row.ZTag, ReservationKind.ZTag, equipmentUuid, batch.ClusterId,
+                    firstPublishedBy, nowUtc, resByKey);
+                MergeReservation(row.SAPID, ReservationKind.SAPID, equipmentUuid, batch.ClusterId,
+                    firstPublishedBy, nowUtc, resByKey);
             }
 
-            batch.FinalisedAtUtc = DateTime.UtcNow;
+            batch.FinalisedAtUtc = nowUtc;
-            await db.SaveChangesAsync(ct).ConfigureAwait(false);
+            try
+            {
+                await db.SaveChangesAsync(ct).ConfigureAwait(false);
+            }
+            catch (DbUpdateException ex) when (IsReservationUniquenessViolation(ex))
+            {
+                throw new ExternalIdReservationConflictException(
+                    "Finalise rejected: one or more ZTag/SAPID values were reserved by another operator " +
+                    "between batch preview and commit. Inspect active reservations + retry after resolving the conflict.",
+                    ex);
+            }
             if (tx is not null) await tx.CommitAsync(ct).ConfigureAwait(false);
         }
         catch
@@ -193,6 +232,71 @@ public sealed class EquipmentImportBatchService(OtOpcUaConfigDbContext db)
         }
     }
 
+    /// <summary>
+    ///     Merge one external-ID reservation for an equipment row. Three outcomes:
+    ///     (1) value is empty → skip; (2) reservation exists for same <paramref name="equipmentUuid"/>
+    ///     → bump <c>LastPublishedAt</c>; (3) reservation exists for a different EquipmentUuid
+    ///     → throw <see cref="ExternalIdReservationConflictException"/> with the conflicting UUID
+    ///     so the caller sees which equipment already owns the value; (4) no reservation → create new.
+    /// </summary>
+    private void MergeReservation(
+        string? value,
+        ReservationKind kind,
+        Guid equipmentUuid,
+        string clusterId,
+        string firstPublishedBy,
+        DateTime nowUtc,
+        Dictionary<(ReservationKind, string), ExternalIdReservation> cache)
+    {
+        if (string.IsNullOrWhiteSpace(value)) return;
+
+        var key = (kind, value.ToLowerInvariant());
+        if (cache.TryGetValue(key, out var existing))
+        {
+            if (existing.EquipmentUuid != equipmentUuid)
+                throw new ExternalIdReservationConflictException(
+                    $"{kind} '{value}' is already reserved by EquipmentUuid {existing.EquipmentUuid} " +
+                    $"(first published {existing.FirstPublishedAt:u} on cluster '{existing.ClusterId}'). " +
+                    $"Refusing to re-assign to {equipmentUuid}.");
+
+            existing.LastPublishedAt = nowUtc;
+            return;
+        }
+
+        var fresh = new ExternalIdReservation
+        {
+            ReservationId = Guid.NewGuid(),
+            Kind = kind,
+            Value = value,
+            EquipmentUuid = equipmentUuid,
+            ClusterId = clusterId,
+            FirstPublishedAt = nowUtc,
+            FirstPublishedBy = firstPublishedBy,
+            LastPublishedAt = nowUtc,
+        };
+        db.ExternalIdReservations.Add(fresh);
+        cache[key] = fresh;
+    }
+
+    /// <summary>
+    ///     True when the <see cref="DbUpdateException"/> root-cause was the filtered-unique
+    ///     index <c>UX_ExternalIdReservation_KindValue_Active</c> — i.e. another transaction
+    ///     won the race between our cache-load + commit. SQL Server surfaces this as 2601 / 2627.
+    /// </summary>
+    private static bool IsReservationUniquenessViolation(DbUpdateException ex)
+    {
+        for (Exception? inner = ex; inner is not null; inner = inner.InnerException)
+        {
+            if (inner is Microsoft.Data.SqlClient.SqlException sql &&
+                (sql.Number == 2601 || sql.Number == 2627) &&
+                sql.Message.Contains("UX_ExternalIdReservation_KindValue_Active", StringComparison.OrdinalIgnoreCase))
+            {
+                return true;
+            }
+        }
+        return false;
+    }
+
     /// <summary>List batches created by the given user. Finalised batches are archived; include them on demand.</summary>
     public async Task<IReadOnlyList<EquipmentImportBatch>> ListByUserAsync(string createdBy, bool includeFinalised, CancellationToken ct)
     {
@@ -205,3 +309,16 @@ public sealed class EquipmentImportBatchService(OtOpcUaConfigDbContext db)
 
 public sealed class ImportBatchNotFoundException(string message) : Exception(message);
 public sealed class ImportBatchAlreadyFinalisedException(string message) : Exception(message);
+
+/// <summary>
+///     Thrown when a <c>FinaliseBatchAsync</c> call detects that one of its ZTag/SAPID values is
+///     already reserved by a different EquipmentUuid — either from a prior published generation
+///     or a concurrent finalise that won the race. The operator sees the message + the conflicting
+///     equipment ownership so they can resolve the conflict (pick a new ZTag, release the existing
+///     reservation via <c>sp_ReleaseExternalIdReservation</c>, etc.) and retry the finalise.
+/// </summary>
+public sealed class ExternalIdReservationConflictException : Exception
+{
+    public ExternalIdReservationConflictException(string message) : base(message) { }
+    public ExternalIdReservationConflictException(string message, Exception inner) : base(message, inner) { }
+}

src/ZB.MOM.WW.OtOpcUa.Core/Resilience/AlarmSurfaceInvoker.cs

@@ -0,0 +1,129 @@
+using ZB.MOM.WW.OtOpcUa.Core.Abstractions;
+
+namespace ZB.MOM.WW.OtOpcUa.Core.Resilience;
+
+/// <summary>
+///     Wraps the three mutating surfaces of <see cref="IAlarmSource"/>
+///     (<see cref="IAlarmSource.SubscribeAlarmsAsync"/>, <see cref="IAlarmSource.UnsubscribeAlarmsAsync"/>,
+///     <see cref="IAlarmSource.AcknowledgeAsync"/>) through <see cref="CapabilityInvoker"/> so the
+///     Phase 6.1 resilience pipeline runs — retry semantics match
+///     <see cref="DriverCapability.AlarmSubscribe"/> (retries by default) and
+///     <see cref="DriverCapability.AlarmAcknowledge"/> (does NOT retry per decision #143).
+/// </summary>
+/// <remarks>
+///     <para>Multi-host dispatch: when the driver implements <see cref="IPerCallHostResolver"/>,
+///     each source-node-id is resolved individually + grouped by host so a dead PLC inside a
+///     multi-device driver doesn't poison the sibling hosts' breakers. Drivers with a single
+///     host fall back to <see cref="IDriver.DriverInstanceId"/> as the single-host key.</para>
+///
+///     <para>Why this lives here + not on <see cref="CapabilityInvoker"/>: alarm surfaces have a
+///     handle-returning shape (SubscribeAlarmsAsync returns <see cref="IAlarmSubscriptionHandle"/>)
+///     + a per-call fan-out (AcknowledgeAsync gets a batch of
+///     <see cref="AlarmAcknowledgeRequest"/>s that may span multiple hosts). Keeping the fan-out
+///     logic here keeps the invoker's execute-overloads narrow.</para>
+/// </remarks>
+public sealed class AlarmSurfaceInvoker
+{
+    private readonly CapabilityInvoker _invoker;
+    private readonly IAlarmSource _alarmSource;
+    private readonly IPerCallHostResolver? _hostResolver;
+    private readonly string _defaultHost;
+
+    public AlarmSurfaceInvoker(
+        CapabilityInvoker invoker,
+        IAlarmSource alarmSource,
+        string defaultHost,
+        IPerCallHostResolver? hostResolver = null)
+    {
+        ArgumentNullException.ThrowIfNull(invoker);
+        ArgumentNullException.ThrowIfNull(alarmSource);
+        ArgumentException.ThrowIfNullOrWhiteSpace(defaultHost);
+
+        _invoker = invoker;
+        _alarmSource = alarmSource;
+        _defaultHost = defaultHost;
+        _hostResolver = hostResolver;
+    }
+
+    /// <summary>
+    ///     Subscribe to alarm events for a set of source node ids, fanning out by resolved host
+    ///     so per-host breakers / bulkheads apply. Returns one handle per host — callers that
+    ///     don't care about per-host separation may concatenate them.
+    /// </summary>
+    public async Task<IReadOnlyList<IAlarmSubscriptionHandle>> SubscribeAsync(
+        IReadOnlyList<string> sourceNodeIds,
+        CancellationToken cancellationToken)
+    {
+        ArgumentNullException.ThrowIfNull(sourceNodeIds);
+        if (sourceNodeIds.Count == 0) return [];
+
+        var byHost = GroupByHost(sourceNodeIds);
+        var handles = new List<IAlarmSubscriptionHandle>(byHost.Count);
+        foreach (var (host, ids) in byHost)
+        {
+            var handle = await _invoker.ExecuteAsync(
+                DriverCapability.AlarmSubscribe,
+                host,
+                async ct => await _alarmSource.SubscribeAlarmsAsync(ids, ct).ConfigureAwait(false),
+                cancellationToken).ConfigureAwait(false);
+            handles.Add(handle);
+        }
+        return handles;
+    }
+
+    /// <summary>Cancel an alarm subscription. Routes through the AlarmSubscribe pipeline for parity.</summary>
+    public ValueTask UnsubscribeAsync(IAlarmSubscriptionHandle handle, CancellationToken cancellationToken)
+    {
+        ArgumentNullException.ThrowIfNull(handle);
+        return _invoker.ExecuteAsync(
+            DriverCapability.AlarmSubscribe,
+            _defaultHost,
+            async ct => await _alarmSource.UnsubscribeAlarmsAsync(handle, ct).ConfigureAwait(false),
+            cancellationToken);
+    }
+
+    /// <summary>
+    ///     Acknowledge alarms. Fans out by resolved host; each host's batch runs through the
+    ///     AlarmAcknowledge pipeline (no-retry per decision #143 — an alarm-ack is not idempotent
+    ///     at the plant-floor acknowledgement level even if the OPC UA spec permits re-issue).
+    /// </summary>
+    public async Task AcknowledgeAsync(
+        IReadOnlyList<AlarmAcknowledgeRequest> acknowledgements,
+        CancellationToken cancellationToken)
+    {
+        ArgumentNullException.ThrowIfNull(acknowledgements);
+        if (acknowledgements.Count == 0) return;
+
+        var byHost = _hostResolver is null
+            ? new Dictionary<string, List<AlarmAcknowledgeRequest>> { [_defaultHost] = acknowledgements.ToList() }
+            : acknowledgements
+                .GroupBy(a => _hostResolver.ResolveHost(a.SourceNodeId))
+                .ToDictionary(g => g.Key, g => g.ToList());
+
+        foreach (var (host, batch) in byHost)
+        {
+            var batchSnapshot = batch; // capture for the lambda
+            await _invoker.ExecuteAsync(
+                DriverCapability.AlarmAcknowledge,
+                host,
+                async ct => await _alarmSource.AcknowledgeAsync(batchSnapshot, ct).ConfigureAwait(false),
+                cancellationToken).ConfigureAwait(false);
+        }
+    }
+
+    private Dictionary<string, List<string>> GroupByHost(IReadOnlyList<string> sourceNodeIds)
+    {
+        if (_hostResolver is null)
+            return new Dictionary<string, List<string>> { [_defaultHost] = sourceNodeIds.ToList() };
+
+        var result = new Dictionary<string, List<string>>(StringComparer.Ordinal);
+        foreach (var id in sourceNodeIds)
+        {
+            var host = _hostResolver.ResolveHost(id);
+            if (!result.TryGetValue(host, out var list))
+                result[host] = list = new List<string>();
+            list.Add(id);
+        }
+        return result;
+    }
+}

tests/ZB.MOM.WW.OtOpcUa.Admin.Tests/EquipmentImportBatchServiceTests.cs

@@ -23,11 +23,13 @@ public sealed class EquipmentImportBatchServiceTests : IDisposable
 
     public void Dispose() => _db.Dispose();
 
+    // Unique SAPID per row — FinaliseBatch reserves ZTag + SAPID via filtered-unique index, so
+    // two rows sharing a SAPID under different EquipmentUuids collide as intended.
     private static EquipmentCsvRow Row(string zTag, string name = "eq-1") => new()
     {
         ZTag = zTag,
         MachineCode = "mc",
-        SAPID = "sap",
+        SAPID = $"sap-{zTag}",
         EquipmentId = "eq-id",
         EquipmentUuid = Guid.NewGuid().ToString(),
         Name = name,
@@ -162,4 +164,93 @@ public sealed class EquipmentImportBatchServiceTests : IDisposable
         await _svc.DropBatchAsync(Guid.NewGuid(), CancellationToken.None);
         // no throw
     }
+
+    [Fact]
+    public async Task FinaliseBatch_Creates_ExternalIdReservations_ForZTagAndSAPID()
+    {
+        var batch = await _svc.CreateBatchAsync("c1", "alice", CancellationToken.None);
+        await _svc.StageRowsAsync(batch.Id, [Row("z-new-1")], [], CancellationToken.None);
+
+        await _svc.FinaliseBatchAsync(batch.Id, 1, "drv", "line", CancellationToken.None);
+
+        var active = await _db.ExternalIdReservations.AsNoTracking()
+            .Where(r => r.ReleasedAt == null)
+            .ToListAsync();
+        active.Count.ShouldBe(2);
+        active.ShouldContain(r => r.Kind == ZB.MOM.WW.OtOpcUa.Configuration.Enums.ReservationKind.ZTag && r.Value == "z-new-1");
+        active.ShouldContain(r => r.Kind == ZB.MOM.WW.OtOpcUa.Configuration.Enums.ReservationKind.SAPID && r.Value == "sap-z-new-1");
+    }
+
+    [Fact]
+    public async Task FinaliseBatch_SameEquipmentUuid_ReusesExistingReservation()
+    {
+        var batch1 = await _svc.CreateBatchAsync("c1", "alice", CancellationToken.None);
+        var sharedUuid = Guid.NewGuid();
+        var row = new EquipmentCsvRow
+        {
+            ZTag = "z-shared", MachineCode = "mc", SAPID = "sap-shared",
+            EquipmentId = "eq-1", EquipmentUuid = sharedUuid.ToString(),
+            Name = "eq-1", UnsAreaName = "a", UnsLineName = "l",
+        };
+        await _svc.StageRowsAsync(batch1.Id, [row], [], CancellationToken.None);
+        await _svc.FinaliseBatchAsync(batch1.Id, 1, "drv", "line", CancellationToken.None);
+
+        var countAfterFirst = _db.ExternalIdReservations.Count(r => r.ReleasedAt == null);
+
+        // Second finalise with same EquipmentUuid + same ZTag — should NOT create a duplicate.
+        var batch2 = await _svc.CreateBatchAsync("c1", "alice", CancellationToken.None);
+        await _svc.StageRowsAsync(batch2.Id, [row], [], CancellationToken.None);
+        await _svc.FinaliseBatchAsync(batch2.Id, 2, "drv", "line", CancellationToken.None);
+
+        _db.ExternalIdReservations.Count(r => r.ReleasedAt == null).ShouldBe(countAfterFirst);
+    }
+
+    [Fact]
+    public async Task FinaliseBatch_DifferentEquipmentUuid_SameZTag_Throws_Conflict()
+    {
+        var batchA = await _svc.CreateBatchAsync("c1", "alice", CancellationToken.None);
+        var rowA = new EquipmentCsvRow
+        {
+            ZTag = "z-collide", MachineCode = "mc-a", SAPID = "sap-a",
+            EquipmentId = "eq-a", EquipmentUuid = Guid.NewGuid().ToString(),
+            Name = "a", UnsAreaName = "ar", UnsLineName = "ln",
+        };
+        await _svc.StageRowsAsync(batchA.Id, [rowA], [], CancellationToken.None);
+        await _svc.FinaliseBatchAsync(batchA.Id, 1, "drv", "line", CancellationToken.None);
+
+        var batchB = await _svc.CreateBatchAsync("c1", "bob", CancellationToken.None);
+        var rowB = new EquipmentCsvRow
+        {
+            ZTag = "z-collide", MachineCode = "mc-b", SAPID = "sap-b", // same ZTag, different EquipmentUuid
+            EquipmentId = "eq-b", EquipmentUuid = Guid.NewGuid().ToString(),
+            Name = "b", UnsAreaName = "ar", UnsLineName = "ln",
+        };
+        await _svc.StageRowsAsync(batchB.Id, [rowB], [], CancellationToken.None);
+
+        var ex = await Should.ThrowAsync<ExternalIdReservationConflictException>(() =>
+            _svc.FinaliseBatchAsync(batchB.Id, 2, "drv", "line", CancellationToken.None));
+        ex.Message.ShouldContain("z-collide");
+
+        // Second finalise must have rolled back — no partial Equipment row for batch B.
+        var equipmentB = await _db.Equipment.AsNoTracking()
+            .Where(e => e.EquipmentId == "eq-b")
+            .ToListAsync();
+        equipmentB.ShouldBeEmpty();
+    }
+
+    [Fact]
+    public async Task FinaliseBatch_EmptyZTagAndSAPID_SkipsReservation()
+    {
+        var batch = await _svc.CreateBatchAsync("c1", "alice", CancellationToken.None);
+        var row = new EquipmentCsvRow
+        {
+            ZTag = "", MachineCode = "mc", SAPID = "",
+            EquipmentId = "eq-nil", EquipmentUuid = Guid.NewGuid().ToString(),
+            Name = "nil", UnsAreaName = "ar", UnsLineName = "ln",
+        };
+        await _svc.StageRowsAsync(batch.Id, [row], [], CancellationToken.None);
+        await _svc.FinaliseBatchAsync(batch.Id, 1, "drv", "line", CancellationToken.None);
+
+        _db.ExternalIdReservations.Count().ShouldBe(0);
+    }
 }

tests/ZB.MOM.WW.OtOpcUa.Core.Tests/Resilience/AlarmSurfaceInvokerTests.cs

@@ -0,0 +1,127 @@
+using Shouldly;
+using Xunit;
+using ZB.MOM.WW.OtOpcUa.Core.Abstractions;
+using ZB.MOM.WW.OtOpcUa.Core.Resilience;
+
+namespace ZB.MOM.WW.OtOpcUa.Core.Tests.Resilience;
+
+[Trait("Category", "Unit")]
+public sealed class AlarmSurfaceInvokerTests
+{
+    private static readonly DriverResilienceOptions TierAOptions = new() { Tier = DriverTier.A };
+
+    [Fact]
+    public async Task SubscribeAsync_EmptyList_ReturnsEmpty_WithoutDriverCall()
+    {
+        var driver = new FakeAlarmSource();
+        var surface = NewSurface(driver, defaultHost: "h");
+
+        var handles = await surface.SubscribeAsync([], CancellationToken.None);
+
+        handles.Count.ShouldBe(0);
+        driver.SubscribeCallCount.ShouldBe(0);
+    }
+
+    [Fact]
+    public async Task SubscribeAsync_SingleHost_RoutesThroughDefaultHost()
+    {
+        var driver = new FakeAlarmSource();
+        var surface = NewSurface(driver, defaultHost: "h1");
+
+        var handles = await surface.SubscribeAsync(["src-1", "src-2"], CancellationToken.None);
+
+        handles.Count.ShouldBe(1);
+        driver.SubscribeCallCount.ShouldBe(1);
+        driver.LastSubscribedIds.ShouldBe(["src-1", "src-2"]);
+    }
+
+    [Fact]
+    public async Task SubscribeAsync_MultiHost_FansOutByResolvedHost()
+    {
+        var driver = new FakeAlarmSource();
+        var resolver = new StubResolver(new Dictionary<string, string>
+        {
+            ["src-1"] = "plc-a",
+            ["src-2"] = "plc-b",
+            ["src-3"] = "plc-a",
+        });
+        var surface = NewSurface(driver, defaultHost: "default-ignored", resolver: resolver);
+
+        var handles = await surface.SubscribeAsync(["src-1", "src-2", "src-3"], CancellationToken.None);
+
+        handles.Count.ShouldBe(2);                 // one per distinct host
+        driver.SubscribeCallCount.ShouldBe(2);     // one driver call per host
+    }
+
+    [Fact]
+    public async Task AcknowledgeAsync_DoesNotRetry_OnFailure()
+    {
+        var driver = new FakeAlarmSource { AcknowledgeShouldThrow = true };
+        var surface = NewSurface(driver, defaultHost: "h1");
+
+        await Should.ThrowAsync<InvalidOperationException>(() =>
+            surface.AcknowledgeAsync([new AlarmAcknowledgeRequest("s", "c", null)], CancellationToken.None));
+
+        driver.AcknowledgeCallCount.ShouldBe(1, "AlarmAcknowledge must not retry — decision #143");
+    }
+
+    [Fact]
+    public async Task SubscribeAsync_Retries_Transient_Failures()
+    {
+        var driver = new FakeAlarmSource { SubscribeFailuresBeforeSuccess = 2 };
+        var surface = NewSurface(driver, defaultHost: "h1");
+
+        await surface.SubscribeAsync(["src"], CancellationToken.None);
+
+        driver.SubscribeCallCount.ShouldBe(3, "AlarmSubscribe retries by default — decision #143");
+    }
+
+    private static AlarmSurfaceInvoker NewSurface(
+        IAlarmSource driver,
+        string defaultHost,
+        IPerCallHostResolver? resolver = null)
+    {
+        var builder = new DriverResiliencePipelineBuilder();
+        var invoker = new CapabilityInvoker(builder, "drv-1", () => TierAOptions);
+        return new AlarmSurfaceInvoker(invoker, driver, defaultHost, resolver);
+    }
+
+    private sealed class FakeAlarmSource : IAlarmSource
+    {
+        public int SubscribeCallCount { get; private set; }
+        public int AcknowledgeCallCount { get; private set; }
+        public int SubscribeFailuresBeforeSuccess { get; set; }
+        public bool AcknowledgeShouldThrow { get; set; }
+        public IReadOnlyList<string> LastSubscribedIds { get; private set; } = [];
+
+        public Task<IAlarmSubscriptionHandle> SubscribeAlarmsAsync(
+            IReadOnlyList<string> sourceNodeIds, CancellationToken cancellationToken)
+        {
+            SubscribeCallCount++;
+            LastSubscribedIds = sourceNodeIds;
+            if (SubscribeCallCount <= SubscribeFailuresBeforeSuccess)
+                throw new InvalidOperationException("transient");
+            return Task.FromResult<IAlarmSubscriptionHandle>(new StubHandle($"h-{SubscribeCallCount}"));
+        }
+
+        public Task UnsubscribeAlarmsAsync(IAlarmSubscriptionHandle handle, CancellationToken cancellationToken)
+            => Task.CompletedTask;
+
+        public Task AcknowledgeAsync(
+            IReadOnlyList<AlarmAcknowledgeRequest> acknowledgements, CancellationToken cancellationToken)
+        {
+            AcknowledgeCallCount++;
+            if (AcknowledgeShouldThrow) throw new InvalidOperationException("ack boom");
+            return Task.CompletedTask;
+        }
+
+        public event EventHandler<AlarmEventArgs>? OnAlarmEvent { add { } remove { } }
+    }
+
+    private sealed record StubHandle(string DiagnosticId) : IAlarmSubscriptionHandle;
+
+    private sealed class StubResolver(Dictionary<string, string> map) : IPerCallHostResolver
+    {
+        public string ResolveHost(string fullReference) => map[fullReference];
+    }
+}