Who uses a process per request for serving Python apps? That must be very uncommon. Even if you use a worker pool that isn’t going to restart a whole process just because of an errant exception in a request handler.
Also as noted if your whole process crashes because of errant input to int() you are beyond fucked in other ways.
There are inputs that can slow it down by hours. Maybe the set the limit too low. Maybe they should have instead merged the PR that improves the speed by a huge amount. They didn't.