🐛 [Bug] Rate limiter scans all clients under the mutex on every single request

[PranavAgarkar07]

The Problem

Every rate-limited request triggers a full scan of all 4096 client entries to clean up expired ones — all while holding a mutex. 🐌

This cleanup (prune()) doesn't need to happen on every request. It should run in a background goroutine every few minutes. Currently it serializes all rate-limited operations for no good reason.

Where to Look

• beamsync/server.go — clientRateLimiter.allow() at line 110. The l.prune(now) call at line 120 is inside the lock.
• clientRateLimiter.prune() at line 161 — the full map iteration.
• Move pruning to a background ticker goroutine.

The Goal

• Prune expired entries in a background goroutine, not on every request
• The allow() method should be O(1) — just a map lookup and update
• Test: send requests from 1000 IPs and check lock hold time is sub-millisecond

[mittalsonal]

Hi maintainer, I verified this is still present on current main: clientRateLimiter.allow() calls l.prune(now) while holding l.mu, and prune iterates the full clients map. Please assign this to me for GSSoC. I can move pruning to a periodic background ticker, keep allow() to lookup/update/optional eviction only, and add a regression/performance-oriented test around many client entries.