Never show the customer a big scary error message. In the unlikely event something goes wrong, dispatch an error event so the engineers can track the issue. They’ll pull detailed logs, and know what to do. Meanwhile, guide the customer to return later with the error screen, and collect their userId if possible. Once the issue is fixed, send them an apology email to let them know they should try again
Most companies just follow the first half of that process, unfortunately
I work for a big tech company, and part of our incident response procedure is to identify impacted users to send apology comms after cleaning everything up. Still, running those queries is a pain in the ass, so I understand why most companies don’t do that