Fitbit Air: a valid DTLS key may NOT decrypt the history dump — a separate AES-EAX layer gates sleep/SpO2
Major added difficulty discovered while scoping the Fitbit Air crack: even a cryptographically-proven DTLS-PSK session key does NOT necessarily yield sleep/SpO2/HRV. Reports indicate the band's history/sync dump can be encrypted under a SEPARATE AES-EAX key, distinct from the DTLS session key. So the pipeline Gattlink->DTLS-PSK->CoAP gets you the transport and live-ish endpoints, but the stored measurement dump has its own second encryption layer to defeat. Implication: 'get the DTLS key -> SpO2/sleep works' is wrong twice over (first the key must be real not a key ID, then the dump needs yet another key). Also: an Android EMULATOR is a dead end for the key because it can't BLE-pair with the physical band to make the app fetch that band's key (the Bumble USB-HCI bridge is a deep rabbit hole). Real reference code for the stack exists in the paceaitian/fitbit-air-notification-bridge repo (GoldenGate SetTlsKeyRpc.kt, TlsKeyResolver.swift). Treat full offline sleep/SpO2 extraction as a serious multi-gate research project, not a hackathon deliverable.