UnitTestBridgeTests hangs forever on AgentCommandTests, silently skipping 12 of 34 bridge rows #489

Ashlar.Tests.CLI.Framework.UnitTestBridgeTests.Framework_unit_test_passes(testType: AgentCommandTests) does not fail — it hangs forever.

Measured with --blame-hang: the Sequence xml marks that row Completed="False", and 12 of the 34 bridge rows never execute as a result. It still hangs with ASHLAR_ALLOW_MOCK=1 set, so this is not the usual "no reachable model provider" refusal.

Consequences:

• Any CI lane that runs Ashlar.Tests.CLI unfiltered will hang rather than fail. This is why the round-3 work had to explicitly exclude UnitTestBridgeTests when repairing the stale tier filters — a hanging gate is worse than the vacuous pass it replaced.
• The 12 unexecuted rows are silently uncovered; nobody is watching them.

Worth finding whether AgentCommandTests is genuinely deadlocking or waiting on something with no timeout, and giving the bridge a bounded timeout either way so a hang reports as a failure. A test framework that can hang the suite it is bridging defeats the point of the bridge.

Found by the round-3 adversarial review's verification agent, which checked rather than accepted the earlier claim that this was a known environmental flake.