What this tests

timeoutsparsing

Point your client at this specimen and find out — before prod does — whether it has a timeout, and whether the timeout has a timeout.

Anatomy

The response begins — status line, part of a header — and then the connection closes. "Content-T" and then nothing, forever. Not a truncated body; a truncated envelope, the TCP equivalent of a letter arriving with the corner torn off before you can see who it's from. This is what a backend crash between "headers committed" and "headers sent" looks like from the wire's perspective — the proxy got half a response and the rest is in the postmortem. Happens at the seam between frameworks and servers, which is why it never happens in your test environment. Clients that read until connection close get a parse error at a byte offset, not a status code. Clients that wait for complete headers get a timeout that attributes itself to the network. Either way, the answer is the same: the response never completed, and your error handling has to know that "never completed" is a status too.

Replay it

curl -i https://faultydata.com:8443/r/FD-2026-081

This exhibit breaks the connection itself, so replay it on the direct line (port 8443), where nothing sits between you and the bytes. On the main site the web server in front tidies the wreckage into a polite 502, and the button below will show you exactly that.

// response appears here

Signature

{Status:0 Headers:map[] Body: Corrupt: Delay:<nil> Flaky:<nil> Sequence:<nil> Drip:<nil> StallAfter:0 Hang:false Reset:false Raw:HTTP/1.1 200 OK
Content-T Phases:[]}

Keep digging

FD-2026-070 — Content-Length lies (body shorter than declared) FD-2026-084 — Response body larger than expected (2MB JSON) FD-2026-008 — WAF challenge page instead of JSON (Cloudflare pattern) FD-2026-019 — JSON body, text/plain label FD-2026-041 — Retry-After as HTTP-date FD-2026-071 — 204 No Content... with content