Ah, I remember being called away from my new year holiday when an engineer dropped our entire database.

This happened because they didn't realise they were connected to the production database (rather than their local dev instance). We were a business intelligence company, so that data was vital. Luckily we had a analysis cluster we could restore from, but afterwards I ensured that backups were happening... never again.

(Why were the backups not already set up? Because they were not trivial due to the size of the cluster and having only been CTO for a few months there was a long list of things that were urgently needed)