Requests queue in connection backlog when workers are saturated

When all Gunicorn workers are busy, incoming requests queue in the connection backlog, causing end-to-end response times (measured at Nginx) to significantly exceed Gunicorn's internal processing time. Example: Nginx reported 3-22 second response times while Gunicorn processed requests in under 1 second.

Increase the number of Gunicorn workers by adjusting the workers configuration parameter. Monitor the difference between Nginx upstream_response_time and Gunicorn's internal request duration to detect queueing. Implement busy worker tracking to determine optimal worker count and detect saturation proactively.