What this tests
headersparsing
Point your client at this specimen and find out — before prod does — whether it reads the headers, or just the status code.
Anatomy
The headers swear the body is gzip; the body is, in fact, plaintext. Some middlebox stripped the compression and left the label — the courier removed the goods and kept the "FRAGILE" sticker, presumably because the sticker was doing all the structural work anyway. Clients that respect Content-Encoding feed plaintext into zlib and get a decompression error that mentions nothing about the actual crime, which is upstream and unaccountable as ever. Clients that ignore headers entirely and sniff the bytes are accidentally correct — a rare instance where doing less work is the right answer, which is why it's so rarely tried. Deceptively common in the wild: anywhere a proxy, a CDN, a service mesh sidecar, or a well-meaning compression middleware has ever touched a response. Which is to say, everywhere.
Replay it
curl -i https://faultydata.com/r/FD-2026-036
// response appears hereSignature
{Status:200 Headers:map[Content-Encoding:gzip] Body: Corrupt: Delay:<nil> Flaky:<nil> Sequence:<nil> Drip:<nil> StallAfter:0 Hang:false Reset:false Raw: Phases:[]}
Keep digging
FD-2026-039 — XML where JSON was promised FD-2026-052 — JSONP callback wrapper FD-2026-058 — CSV when JSON was expected FD-2026-019 — JSON body, text/plain label FD-2026-041 — Retry-After as HTTP-date FD-2026-086 — charset=utf-16 lie on a UTF-8 body