Monitord replaying does not really work with MySQL as it does not clean out previous data in the database. The user will have to drop and create the database (or use a new one...). The problem is that with MySQL, you might have many workflow instances sharing the database, so dropping is not an option. So you are stuck with replaying all workflows, not only the one which had problems.

Note that I have only have a couple of short discussions with the team about this issue, and I might have misunderstood the setup, but...

I'm confused on why the sqlite and mysql are working differently. It seems like when replaying with sqlite, the database is recursively cleaned, but with mysql nothing is cleaned. This seems backwards as sqlite is basically the local mode, not shared with other workflows and can be fully deleted before replay, while mysql is the shared one and should be "carefully" cleaned.