The merge of WT-16221 caused failures in both:

• test/format multi-node runs (verify mirror tables fails and a cursor used in verification references a dhandle that has been freed) 
• test/model runs in eviction thread startup.

Looking at these bugs, it's possible that either could be due to:

• a "too aggressive" expiration of btrees (i.e. a bug in the fix for WT-16221)
• or an "appropriately aggressive" expiration of btrees (no bug in the fix), but this creates a kind of stress testing of sweep that might expose other pre-existing bugs in the system.

At the moment, we don't know enough about this - and we don't know whether the WT-16221 can cause some different instability on DSC standbys or even primaries.

So we're going to revert until we know more.  That will keep our WT tests green in the meantime.