Seeing sporadic timeouts on POST /auth/login since about 14:30 UTC; about 1 in 20 requests return 504 from the edge in us-east-1 when called via curl and our Node client. Can anyone confirm similar behavior or share if forcing HTTP/1.1 vs HTTP/2 changes it on your side?
Saw the same in us-east-1 around 14:30; forcing curl --http1.1 made the edge 504s disappear. For me it was the ALB idle timeout being lower than the Node client’s keep‑alive — bumping the ALB to 60s cleared the “1 in 20” failures. Can you try --http1.1 and share your origin/ALB idle timeout?
Only hit this after about 14:30 UTC when HTTP/2 was negotiated; forcing a real Content-Length (no chunked) on POST /auth/login made the 504s disappear… As a stop‑gap in Node I disabled connection reuse on that call (Connection: close) and it’s been clean since, @darryl_hughes21.
I’ve noticed similar behavior with the edge timeouts too. After about 14:30 UTC, switching to HTTP/1.1 for my POST requests seemed to help. It drives me nuts trying to pin down these intermittent issues — are you also seeing any impacts on user logins during peak times?