when do you actually need a distributed rate limiter tho, like what are the main reasons

hey folks, ive been thinking about rate limiting lately, specifically for apis exposed to external consumers. weve handled simple rate limiting at the api gateway level for a while, but with microservices, things get more complex. when do you guys typically decide its time to move beyond gateway-level limiting and implement a dedicated distributed rate limiting solution. what are the key triggers or pain points that signal this need