Many server machines are actually running less powerful hardware than your dev machine, and are always handling more requests in parallel. A long enough int to stretch a ~10ms request into a ~100ms request can reduce throughput for all users to 10% [1], possibly more if you’re also logging details about failed requests (which presumably will be the result of such a request).
This may be true, but as we’ve seen from the examples above, the people who are unhappy about the limit won’t be happy until it’s completely gone. So it may as well be low enough that they’ll discover and disable it for their apps, and those of us who mostly live in the “ints never exceed 64 bits” world dominated by other languages don’t have any worries at all about users exceeding reasonable limits.
Edit I made the “10%” up. Let’s just go with “drastically”, because I can’t be bothered doing the actual throughput calculation. Certainly one request taking 10x longer on the CPU is going to use up resources that are needed to process other requests, and cause congestion on the server. ↩︎