During replica set step-down, WT receives a step-down timestamp before MongoDB necessarily completes the transition to secondary. Subsequent writes are directed to ingest tables after the step-down timestamp is set.

Currently, MongoDB cannot cancel this state and resume normal leader operation. If step-down cannot complete, the expected behavior is to terminate the process. Remaining in this state indefinitely is not viable because ingest writes are limited by the WT cache (ingest content is unevictable in this state).

Investigate whether WT can safely cancel the step-down and remain leader. Mirror writes (writing to both stable and ingest) may make it possible to discard the ingest constituent and continue using stable, but the correctness and synchronization requirements need to be established.

The investigation should cover:

• When cancelling step-down is safe.
• How post-step-down-timestamp writes should be handled.
• Required server/WT API and state changes.

Acceptance criteria:

• Document the current behavior.
• Propose a safe design for remaining leader, or explain why it is not feasible.
• Identify required API changes, metrics, and tests.