Summary

DRIVERS-2884 introduced logic to retry operations that fail while attempting to drain the response from a connection that experienced a socket timeout. We should add tests to ensure that we actually (1) retry in such cases and (2) that this retry goes through the appropriate server-selection layer.

See PR comment for more context.

Motivation

Testing completeness

Acceptance Criteria

We should duplicate the two tests that in client-side-operation-timeouts/tests/pending-response.yml that check for case (1):

• write op retries when connection closes server-side while draining response
• read op retries when connection closes server-side while draining response

And extend them to use logging to confirm that a successful retry doesn't happen at the pooling layer:

          - connectionPendingResponseStarted: {} # Pending read fails on first find
          - connectionPendingResponseFailed:
              reason: error
          - connectionClosedEvent:
              reason: error
          - serverSelectionStarted
          - serverSelectionFinished
          - connectionCheckedOutEvent: {}
          - connectionCheckedInEvent: {} # Find finishes.

This has to be done with logging since event monitoring doesn't include server selection.