## Problem

After every LLM call, the rate limiter unconditionally writes counters to Redis even for API keys, users, or teams that have no rate limits configured. These Redis writes are wasted work: the counters are never read for enforcement because there are no limits to enforce. At high throughput this generates unnecessary Redis load and latency on every request.

## Proposed fix

Before the post-call Redis increment, check whether the relevant key/user/team has any rate limit configured. If none is set, skip the write entirely. This is safe because the counter value is only meaningful when there is a limit to compare it against.

## Files

- `litellm/proxy/hooks/` or `litellm/proxy/utils.py` (rate limiter post-call logic)