fix: add --preload to Gunicorn to prevent worker timeout on startup

Without --preload, each Gunicorn sync worker independently loads the
Django WSGI application after forking. During this loading phase, the
worker does not send heartbeats to the arbiter. On resource-constrained
environments (staging, small Docker containers), this loading can exceed
the 120-second timeout, causing the arbiter to kill the worker with
WORKER TIMEOUT before it ever serves a request.

With --preload, the application is loaded once in the master process
before forking workers. Workers start with the app already loaded,
eliminating the startup timeout risk.

https://claude.ai/code/session_01AztUaxUwQFtPfzRVyDzeNz

Author: claude

Dockerfile

@@ -29,5 +29,6 @@ CMD ["gunicorn", "core.wsgi:application", \
      "--worker-tmp-dir", "/dev/shm", \
      "--workers", "3", \
      "--timeout", "120", \
+     "--preload", \
      "--forwarded-allow-ips", "*", \
      "--access-logfile", "-"]