Investigate AI-agent analysis report on the change-streams 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

• Drivers MUST apply timeouts to change stream establishment, iteration, and resume attempts (§Timeouts) — no
  unified test for timeout enforcement
• Drivers MUST NOT err when they encounter a new operationType (§Response Format) — no forward-compatibility test
  for unknown operation types
• Drivers MUST NOT err when extra fields are encountered in the ns document (§Response Format)
• A driver MUST only attempt to resume once from a resumable error (§Resume Process) — prose test 3 exists but
  limited automated coverage
• Drivers SHOULD attempt to kill the cursor on the server during the resume process (§Resume Process)
• getResumeToken MUST return postBatchResumeToken when batch is empty or exhausted (Prose test 11) — not
  automated
• ChangeStream MUST throw an exception if the server response is missing the resume token (Prose test 2) — prose
  only

Ambiguities

• Resumable error codes list for wire version < 9: It's unclear whether all listed codes (including 63
  StaleShardVersion, 13388 StaleConfig) are truly resumable or contextual. "An error on an aggregate command is not a
  resumable error" is ambiguous for sharded clusters.
• startAfter condition "yet to return a result document": "yet to return" is temporally vague — does it mean
  "never returned" or "in the process of returning"?
• Last document in batch with postBatchResumeToken: If a batch has only one document and postBatchResumeToken
  is present, is that document "the last document"? The spec's wording could be interpreted either way.
• showExpandedEvents server version: ">6.0.0" lacks precision; field visibility changes between 6.0 and 8.2
  create multiple API surfaces
  (DRIVERS-3278 implementing this adjustment).

Inconsistencies

• §Timeouts vs. §Resumable Error: If a timeout fires during a change stream operation, is it an error on aggregate
  or getMore? The spec says "An error on an aggregate command is not a resumable error" but doesn't clarify timeout
  interaction.
• §Resume Process vs. Prose test 3: Prose test 3 says "resume one time" but §Resume Process specifies conditional
  logic (startAfter with no results vs. after results). The unified tests have limited coverage of this branching.
• change-streams-errors.yml vs. spec error code list: Tests only cover a subset of the 17+ error codes listed in
  the spec.

Notes

• 9 unified test files (~18 test cases). Prose tests (19 items) cover advanced resume scenarios but require manual
  implementation.
• Recent changes (2025-03-31) for server 8.2+ field visibility (operationDescription, nsType,
  collectionUUID) may not be fully covered by existing tests
  (DRIVERS-3278,
  DRIVERS-3458 implementing).