Investigate AI-agent analysis report on the unified-test-format specification. This ticket tracks findings identified during a systematic review of the spec: missing test coverage, ambiguous normative language, and spec/test inconsistencies.

Missing Tests

• Entity map concurrent access under contention — spec requires thread-safe access but only basic thread operations
  are tested; race conditions around creation/retrieval not tested
• saveResultAsEntity error when saving a non-BSON driver-level object (e.g., Cursor) to entity map in
  BSON-restricted mode
• $$placeholder substitution failure: missing environment variable or config — behavior not specified
• runOnRequirements with conflicting conditions (e.g., minServerVersion: 5.0 AND maxServerVersion: 4.4) —
  error handling
• schemaVersion with unsupported patch version (e.g., 1.0.1 when runner supports only 1.0.0) — should error or
  pass?
• expectError with multiple optional fields where some match and some don't (e.g., errorCode matches but
  errorCodeName doesn't)
• cursor.next() / cursor.toArray() called on an already-exhausted cursor
• topologyDescription entity snapshot timing: does it capture state before or after the operation?

Ambiguities

• Entity uniqueness enforcement error types: "Uniqueness enforcement" and "referential integrity" required but
  exception types and messages not specified — inconsistent across drivers.
• createChangeStream vs. watch(): "SHOULD NOT use watch()" — what implementation variations exist? Why is
  watch() discouraged?
• serverParameters cache scope: "Result SHOULD be cached" — per test, per test file, or per runner session? TTL
  not specified.
• expectError mutual exclusivity: Spec lists it as mutually exclusive with expectResult and saveResultAsEntity
  but doesn't clarify which takes precedence if multiple are specified.

Inconsistencies

• comment field in event matching: Optional in command monitoring events but spec doesn't explain
  filtering/matching logic when comment is a complex BSON type vs. string.
• ignoreExtraEvents: true semantics: "Events after all specified have matched MUST NOT cause a test failure" —
  does event ordering still matter for the matched events when extra events are interspersed?
• Timeout handling delegation: Timeout semantics for explicit operations vs. cursor iteration differ; spec delegates
  to CSOT spec but unified format doesn't consistently specify which applies in which context.

Notes

• ~638 test files. Comprehensive coverage of happy-path test execution, entity creation, and event assertion.
• Error cases and edge cases in entity map management (concurrency, invalid types) are minimal.
• serverParameters caching scope is under-specified — different drivers may cache at different granularities.
• No ## Changelog section — spec has a changelog but it's embedded inline.