fix(api): set gunicorn keep-alive above the load balancer idle timeout to stop 502s (#11647)

api/src/backend/config/guniconf.py

@@ -56,6 +56,12 @@ preload_app = not DEBUG
 # that may take longer, such as complex API operations.
 timeout = env.int("GUNICORN_TIMEOUT", default=120)
 
+# HTTP keep-alive idle timeout. Must exceed the idle timeout of the proxy or load
+# balancer in front of gunicorn, or it reuses a connection gunicorn just closed
+# and returns a 502. Default clears the common 60s; raise `GUNICORN_KEEPALIVE` to
+# stay above a longer one.
+keepalive = env.int("GUNICORN_KEEPALIVE", default=75)
+
 # Logging
 logconfig_dict = DJANGO_LOGGERS
 gunicorn_logger = logging.getLogger(BackendLogger.GUNICORN)