WT currently doesn’t support graceful step-downs. Reconfiguring WT directly can cause issues, so we restart the connection when switching from leader to follower (in  our internal testing like - test/format). I think mongo does the same.

Skipping the restart may lead to problems due to WT’s internal state not resetting cleanly. One known (eviction) issue is linked in this ticket, but there may be others.

Scope

• Investigate potential issues that arise when not restarting WT on step-down (also tracking how mongodb is handling this would be good).
• List of identified issues

Success Criteria

• All known and newly found issues are documented to be fixed in seperate tickets.