Worker memory leaks cause restarts and task failures

Worker memory usage exceeds 2GB threshold, causing OOM kills and task failures. Memory leaks accumulate over time, particularly in image processing tasks. Worker restarts occurring 3+ times in 1 hour indicate memory issues.

Monitor memory usage per worker with <1GB normal threshold, restart workers when >2GB. Implement worker restart policy after every 500 tasks to prevent leak accumulation. Identify problematic tasks causing 85% of failures. Increase worker RAM allocation if repeated restarts occur.