The schema_abort test and tiered storage does not run long before hitting failures. One type of issue I've seen frequently is that the internal work thread is finding and using stale work units after tiered tables have been dropped or closed/swept.

The WT_TIERED_WORK_UNIT structure currently contains a field WT_TIERED *tiered. Having a work unit store that pointer is unsafe as sweep can and will discard and free that memory.

One potential solution would be for the work unit to store a copy of the URI name and then use session_get_dhandle and session_release_dhandle locally.