What this tests
timeouts
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 TCP handshake completes beautifully and then — silence. No response, no reset, no closure. Just a socket held open like a door in a horror film, with something on the other side that has simply decided not to speak. The connection is alive. The conversation is not. It is the HTTP equivalent of being left on read by a load balancer. Clients with only a connect timeout — a surprisingly popular configuration, usually written by someone who read half a blog post — will wait here until the heat death of the universe or OOM, whichever arrives first. Both are scheduled. Neither will send a notification. Server-capped at sixty seconds, which is still longer than most incident calls and roughly the attention span of a goat at a corporate mindfulness retreat. If your client gives up before the server does, it passes. If it out-waits the server, someone needs to have a quiet word with whoever owns the timeout budget.
Replay it
curl -i https://faultydata.com/r/FD-2026-017
// response appears hereSignature
{Status:0 Headers:map[] Body: Corrupt: Delay:<nil> Flaky:<nil> Sequence:<nil> Drip:<nil> StallAfter:0 Hang:true Reset:false Raw: Phases:[]}
Keep digging
FD-2026-006 — Total reachability loss (Meta BGP withdrawal, Oct 2021) FD-2026-016 — Slow but correct response FD-2026-018 — Partial body then stall FD-2026-025 — Edge DNS unreachable (Akamai, Jul 2021) FD-2026-043 — Datacenter literally on fire (OVH, Mar 2021) FD-2026-057 — Drip-fed response body