review crash recovery scenarios relating to file creation and renames #13043

This issue relates to a comment in code being refactored here: #12967

The comment seems to be overly specific about one of the specific scenarios where a crash can cause trouble.
In general a crash after any update to persistent state is something that should be handled.
This bug^H^H^Hissue is to look at this logic and ensure that testing covers crash recovery after every persistent update in the code path above.