How to ignore all exceptions and keep running at all costs?

Even if some API or some third party service or whatever is bugging out, as a policy I would like the server to do it's best and keep doing whatever is working.

By default, when an exception occurs, it's game over, server down, everything stops working.

What is the recommended way to maybe just log ALL possible exceptions (even typically server life-threatening ones) and just stay alive as if nothing is wrong?

Of course, this would have to be coupled with monitoring, downtime alerts etc. but I think this is always preferable to having the server die.

No file, test, or entry point is named. Start by locating Ktor's server exception handling and logging paths, then determine how failures from APIs and third-party services should be isolated without stopping unrelated work. Done should include a documented policy for which exceptions are caught, logged, and allowed to terminate the server.

Written by the indexing model from the issue text.