The Gunicorn Flask server on Heroku terminates the last request before restarting the worker, resulting in a 503 error: “Connection closed without response.”

I have analyzed hundreds of such requests, and whenever the server encounters the “Connection closed without response” error, it always occurs on the last call before that particular worker restarts.

i have set the following setting on Gunicorn server:

web: gunicorn main:app --threads=3 --max-requests=1200

with 4 workers in each dyno.

gunicorn==22.0.0

Its worth noting that the last call that ends up in error had always zero service time.

I have updated to latest gunicorn version but it had not effect
tried increasing and decreasing the threads + workers as well
reduced latency of database too