Well.. I created a fork bomb on AWS Lambda. It was supposed to check a condition before self executing with a new payload, but the check was borked, so it just always self executed.
It turns out, if you do this _directly_, and not through a second service like Event Bridge, AWS will notice this fact, and will just suspend your lambda for a few minutes until all the executions die out.
It was only a $20 mistake in the end.
The real nightmare with Event Bridge is the default "retry" threshold is 185 times. It's a nuclear level nightmare. I just use the cron part now.