Hey everyone, I’ve been running into a stubborn issue with my Caddy reverse proxy setup over the past couple of days and could really use another set of eyes. I have Caddy configured in front of a standard backend container handling media asset delivery and package downloads. When traffic spikes and users request larger files across cepcutapk, Caddy randomly returns 502 Bad Gateway responses and occasionally hangs on the initial TLS handshake for incoming mobile clients. The backend itself does not seem starved for CPU or memory, and local curl checks to the upstream port respond immediately without errors.
I suspect it might be related to keepalive timeouts or upstream connection pooling in my Caddyfile, especially since chunked responses get abruptly severed after about thirty seconds of sustained transfer. I have tried tweaking the transport settings and dialing timeout parameters, but the handshake latency persists whenever several simultaneous connections open up. Has anyone dealt with similar connection dropouts when serving heavy binary streams through Caddy’s reverse_proxy directive, and is there a specific buffer or transport configuration I should be adjusting to keep these connections stable?