Master process disk I/O blocking causes multi-second response latency

When disk I/O wait spikes occur (particularly on AWS magnetic EBS volumes), Gunicorn master process blocks on filesystem operations, causing multi-second response latency spikes even though worker processes are idle and waiting normally for requests. Correlation observed between disk I/O wait and upstream response time.

Configure worker_tmp_dir to use tmpfs filesystem to avoid disk blocking. Migrate from AWS magnetic EBS to SSD-backed volumes (gp2/gp3 or provisioned IOPS) for predictable performance. Monitor disk I/O wait metrics alongside gunicorn.request.duration. Investigate master process activity during worker lifecycle events.