The 20210405-update-datetime-columns.sql migration in the server_core repo has been consistently failing for several reasons:

• Lock conflicts with jobs executed by the circulation webapp and script runner
• Once those were resolved, it ate all the available disk space on the QA circulation database's RDS instance
• Once the disk space was increased, it failed due to internal deadlocks

The migration script used to run DB migrations on application startup will, by default, wrap the entirety of an SQL-based migration in a single transaction. That is probably responsible for some of the issues.