The intent of this change is to add a fail point in reconciliation that returns an error during __wt_hs_insert_updates the fail point should occur with a low probability and only be enabled in diagnostic builds.
Another question is whether it should also fail during checkpoints or not.
There are two potential fail points:
- In the middle of inserting HS updates
- After inserting all history store updates
The fail point functionality should be built into the timing stress functionality and additionally test format and test checkpoint should be updated to run with them.