ChunkGenerator

• Memory pressure circuit breaker: the throttle scales down once heap usage crosses memory-backoff-pct of max heap and pauses new chunk submissions entirely past memory-pause-pct, resuming automatically once pressure clears.
• Auto-scaled defaults: the inflight ceiling and start target now derive from availableProcessors() and maxMemory() at startup, so the plugin saturates Paper's chunk pipeline on any host without manual tuning.

Changed

• New throttle model: the controller now targets a count of concurrent chunk requests in flight instead of a fixed chunks-per-tick budget. This matches Paper's asynchronous chunk system and removes the budget-loss bug that capped throughput every time the previous hardcoded MAX_INFLIGHT = 200 ceiling was reached.
• Speed readout labelled chunk/s instead of c/s everywhere (BossBar, ActionBar, console log, /cg status, /cg list).

Removed

• Config keys throttle.max-chunks-per-tick and throttle.min-chunks-per-tick (replaced by throttle.max-inflight, throttle.min-inflight and throttle.start-inflight), and the hardcoded MAX_INFLIGHT = 200 ceiling on concurrent chunk loads.

Migration

Existing config.yml files keep working — the new keys fall back to sensible auto-scaled values. To pick them up explicitly, delete plugins/ChunkGenerator/config.yml and the plugin will write a fresh one on the next start.