rate limiting strategies, per user versus global. dont you think we should look at that?

Hey everyone, I'm trying to figure out the best way to do rate limiting for a new service we're building. We're going back and forth on whether to do per-user limits, like 100 requests per minute per API key, or just a global limit for the whole service, maybe 10,000 requests per minute. Both have their ups and downs. Per-user feels more fair, but it's definitely more complicated to set up. Global is simpler, sure, but then one really heavy user could mess things up for everyone else. What have you guys done? When did you pick one over the other, and what made you decide?