fix(opcuaclient): #910 name the per-endpoint connect budget; widen it in the #816 host-name tests

Cause: the client-side PerEndpointConnectTimeout (3 s), not opc-plc. The
validate:null case is the first test the assembly runs, so it pays the
cold-process cost of the first Basic256Sha256 connect: loading the fresh
throwaway certificate, validating the server's, and the asymmetric handshake.
In the G2 full-solution run that local work used up the budget. The opc-plc
log at 13:13:50-53Z shows the discovery channel closing at 50.709, then nothing
from the client for 3 s, then OpenSecureChannel answered at 53.775 and the
client closing the channel 48 ms later. The SDK reports a wait cancelled by
that budget as BadRequestInterrupted. opc-plc auto-accepted the certificate
(--aa, "Trusting certificate ... command line option") with no reject and no
store round-trip.

- Driver: a per-endpoint budget expiry now surfaces as TimeoutException naming
  PerEndpointConnectTimeout instead of BadRequestInterrupted or a bare
  TaskCanceledException. There is no behaviour change. Unit test added
  against a silent TCP listener.
- Test: the ServerHostNameValidationTests secure options use a 15 s budget,
  the same as the browser integration tests' 10 s. Connect latency is not
  what these tests check.
- docs/drivers/OpcUaClient.md: documents what the budget covers.

docs/drivers/OpcUaClient.md

@@ -75,7 +75,10 @@ measurements; they are not fresh SDK 2.0.1 interoperability evidence.
 `EndpointUrls` (a list) takes precedence over the single-URL `EndpointUrl` and
 provides ordered **failover** — the driver tries each candidate in turn at every
 `InitializeAsync` (including the comms-loss watchdog's re-initialise), and the first to connect wins (e.g. a hot-standby pair on
-4840 / 4841). See
+4840 / 4841). Each candidate gets `PerEndpointConnectTimeout` (default 3 s) for discovery + secure channel + session. That
+budget includes **client-side** work (certificate load/validation, the asymmetric handshake), so a cold process on a busy
+host can exceed it against a healthy server; an expiry is reported as `TimeoutException: … PerEndpointConnectTimeout (3 s)`
+(#910 — the SDK alone says `BadRequestInterrupted`, which reads like a server fault). See
 `src/Drivers/ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Contracts/OpcUaClientDriverOptions.cs`
 for every field (security policy/mode, auth type, session timeout, keep-alive,
 browse root, node/depth caps). `ReconnectPeriod` was **deleted** (#676 — unused since #670 removed the SDK reconnect

src/Drivers/ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient/OpcUaClientDriver.cs

@@ -1053,8 +1053,33 @@ public sealed class OpcUaClientDriver : IDriver, ITagDiscovery, IReadable, IWrit
         using var cts = CancellationTokenSource.CreateLinkedTokenSource(ct);
         cts.CancelAfter(_options.PerEndpointConnectTimeout);
 
+        try
+        {
+            return await OpenSessionWithinBudgetAsync(appConfig, endpointUrl, policy, mode, identity, cts.Token)
+                .ConfigureAwait(false);
+        }
+        catch (Exception ex) when (cts.IsCancellationRequested && !ct.IsCancellationRequested)
+        {
+            // Gitea #910: the SDK reports a wait cut short by this budget as BadRequestInterrupted (or a bare
+            // TaskCanceledException), which reads like a server or network fault. Name the budget instead. It also
+            // covers CLIENT-side work (loading the application certificate, validating the server's, the asymmetric
+            // crypto of OpenSecureChannel), so a cold process on a busy host can exceed it with a healthy server.
+            throw new TimeoutException(
+                "The connect did not complete within PerEndpointConnectTimeout " +
+                $"({_options.PerEndpointConnectTimeout.TotalSeconds:0.###} s); the SDK reported: {ex.Message}", ex);
+        }
+    }
+
+    private async Task<ISession> OpenSessionWithinBudgetAsync(
+        ApplicationConfiguration appConfig,
+        string endpointUrl,
+        OpcUaSecurityPolicy policy,
+        OpcUaSecurityMode mode,
+        UserIdentity identity,
+        CancellationToken budget)
+    {
         var selected = await SelectMatchingEndpointAsync(
-            appConfig, endpointUrl, policy, mode, cts.Token).ConfigureAwait(false);
+            appConfig, endpointUrl, policy, mode, budget).ConfigureAwait(false);
         var endpointConfig = EndpointConfiguration.Create(appConfig);
         endpointConfig.OperationTimeout = (int)_options.Timeout.TotalMilliseconds;
         var endpoint = new ConfiguredEndpoint(null, selected, endpointConfig);
@@ -1073,7 +1098,7 @@ public sealed class OpcUaClientDriver : IDriver, ITagDiscovery, IReadable, IWrit
             (uint)_options.SessionTimeout.TotalMilliseconds,
             identity,
             default,                                            // preferredLocales
-            cts.Token).ConfigureAwait(false);
+            budget).ConfigureAwait(false);
 
         session.KeepAliveInterval = (int)_options.KeepAliveInterval.TotalMilliseconds;
         return session;

tests/Drivers/ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.IntegrationTests/ServerHostNameValidationTests.cs

@@ -35,6 +35,12 @@ public sealed class ServerHostNameValidationTests(OpcPlcFixture sim) : IDisposab
         SecurityMode = OpcUaSecurityMode.SignAndEncrypt,
         AutoAcceptCertificates = false,
         CertificateValidation = new OpcUaCertificateValidationOptions { ValidateServerHostName = validateHostName },
+        // Gitea #910: the default 3 s per-endpoint budget covers client-side work too — loading the fresh throwaway
+        // certificate, validating the server's, the first Basic256Sha256 handshake in a cold test process. The first
+        // test in this assembly (the validate:null case) paid that cold cost, and in the full-solution run it ran out:
+        // opc-plc answered OpenSecureChannel at once, but the client reached it only after 3 s of local work and
+        // cancelled 48 ms later (BadRequestInterrupted). What is under test is the host check, not connect latency.
+        PerEndpointConnectTimeout = TimeSpan.FromSeconds(15),
     };
 
     /// <summary>Trust the fixture's server certificate (as advertised on its secure endpoint) in the throwaway PKI.</summary>

tests/Drivers/ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Tests/OpcUaClientFailoverTests.cs

@@ -83,4 +83,30 @@ public sealed class OpcUaClientFailoverTests
         ex.Message.ShouldContain("127.0.0.1:3");
         drv.GetHealth().State.ShouldBe(DriverState.Faulted);
     }
+
+    /// <summary>
+    ///     Gitea #910: an endpoint that accepts TCP but never answers runs out the per-endpoint budget. The SDK reports
+    ///     that as BadRequestInterrupted / a bare cancellation; the failover aggregate must name the budget instead.
+    /// </summary>
+    [Fact]
+    public async Task A_per_endpoint_budget_expiry_is_reported_as_that_budget()
+    {
+        using var silent = new System.Net.Sockets.TcpListener(System.Net.IPAddress.Loopback, 0);
+        silent.Start(); // accepts (kernel backlog) but never reads or replies
+        var port = ((System.Net.IPEndPoint)silent.LocalEndpoint).Port;
+        var opts = new OpcUaClientDriverOptions
+        {
+            EndpointUrl = $"opc.tcp://127.0.0.1:{port}",
+            PerEndpointConnectTimeout = TimeSpan.FromMilliseconds(400),
+            Timeout = TimeSpan.FromSeconds(30),
+            AutoAcceptCertificates = true,
+        };
+        using var drv = new OpcUaClientDriver(opts, "opcua-budget");
+
+        var ex = await Should.ThrowAsync<AggregateException>(async () =>
+            await drv.InitializeAsync("{}", TestContext.Current.CancellationToken));
+
+        ex.Message.ShouldContain("TimeoutException");
+        ex.Message.ShouldContain("PerEndpointConnectTimeout (0.4 s)");
+    }
 }