Question
Your mobile API moved all client traffic onto a single multiplexed HTTP/2 connection per client (over TLS/TCP) to cut handshakes. It works great on good networks. But users on lossy mobile networks now report that when ANY request stalls, ALL their in-flight requests stall together and then recover together — even small independent calls. Server-side everything is healthy and fast; the server's per-stream metrics show no slow handlers. Packet-capture from affected clients shows TCP retransmissions right before each multi-stream stall. How do you triage and what are the options?
Stop the bleeding first (mitigate), then form hypotheses from real signals. Separate root cause from symptom, communicate status as you go, and close with what prevents a repeat.