[matsvarn]

Why

A failed Gen5 GET_DATA_RANGE reply can keep a plausible body from an earlier successful read, including revision byte 1. The current decoder checks only the revision byte. It therefore reports stale range_oldest/range_newest, trim_ts/current_read_ts, and pages_behind (including raw_old_page/read_page) as current values.

The app's single GET_DATA_RANGE consumer (ble_engine.dart) does not check the outer status before it adopts these fields as the backlog target and writes the band_backlog ledger row. A stale read cursor makes the band's drain position look stalled or rewound. That would corrupt the cursor-movement evidence the interrupted-sync trial relies on.

Change

• This ports the behavior of OpenStrap/protocol 11919c3 (gate get_data_range on outer cmd status OpenStrap/protocol#55) to the fork's decoder, which carries the read-cursor fields from 0df79ff.
• On Gen5, both decode blocks now require outer status 1. Together they cover every field this opcode emits.
• Gen4 stays ungated because its status-byte semantics are unconfirmed. This matches the existing GET_CLOCK policy.
• The gate has the same shape as upstream's, to keep a later upstream sync conflict-free.

Verification

Check	Result
New Gen5 status 0/2/3 test on the original decoder (0df79ff)	Failed as expected: status=0 key=range_oldest was emitted
Same test with this change	Passes. No range, backlog, or cursor key is emitted
Gen5 status 1	All fields are emitted with their expected values, including the cursor fields
Gen4 status 0/1/2/3	Unchanged. All fields are emitted
dart analyze --fatal-infos (Dart 3.11.4)	No issues
dart test (full suite)	616 passed, 4 skipped (the skipped tests need an external capture)
dart format --set-exit-if-changed	Flags 39 files. main at 0df79ff flags the same 39, and nothing flagged is on a line this PR adds

No hardware verification yet. The synthetic replies cover the decoder only. The physical interrupted-sync trial runs after the app pins this commit.

Other upstream-only fixes reviewed (not included)

I checked these against the commands the current app sends. The app frames every Gen5 command through buildCommand(..., session.band), and none of them is needed for the capture path:

• Builder profile params (fix cmdAbortHistorical missing profile param OpenStrap/protocol#62 abort, cmdGetBattery: take a profile like its siblings OpenStrap/protocol#63 battery, fix: wire up gen5 framing on remaining unwired command builders OpenStrap/protocol#65 remaining builders, cmdBuzz: take a profile like its siblings OpenStrap/protocol#67 buzz): latent here. The app doesn't call those convenience builders.
• Status gates for battery pack info (gate battery pack info reply on gen5 cmd_status OpenStrap/protocol#57), advertising name and alarm time (gate GET_ADVERTISING_NAME on cmd_status OpenStrap/protocol#58; the alarm library change landed there, not in gate get_alarm_time on status like getclock/getdatarange OpenStrap/protocol#60), select wrist (gate select_wrist decode on cmd_status OpenStrap/protocol#59), body location (gate GET_BODY_LOCATION_AND_STATUS on cmd_status OpenStrap/protocol#61): these affect battery, name, and alarm state, not history sync. They are a good single follow-up port.
• GET_HELLO profile gate (gate GET_HELLO gen5 decode on profile OpenStrap/protocol#69): affects Gen4 only. Gen5 HELLO is already status-gated here.
• R10/version-10 collision (fix R10 misdecode of short historical records with version byte 10 OpenStrap/protocol#68): the app tries parseGen5Historical before decodeRecord/realtimeRr. This is reachable on Gen5 only if a record isn't identified as Gen5, so it is a follow-up.
• Console-log reassembler (fix console log reassembler eating multi-chunk lines OpenStrap/protocol#56), Garmin/Ultrahuman cleanup (c22925e), WHOOP MG (61349b4): not on this path.

[matsvarn]

Hosted CI on the exact head 4bab607: https://github.com/matsvarn/openband5-protocol/actions/runs/36137606205 passed on Dart 3.5.0 and stable. It was run via workflow_dispatch because the fork recorded no push or pull_request runs; see the note in the handoff.

[matsvarn]

Independent review (Grok 4.7, high reasoning): CLEAN, no findings. It reproduced the base failure (status=0 key=range_oldest) on 0df79ff and confirmed the focused tests pass on 4bab607. It also confirmed profile.isGen5 covers every Gen5-family profile, including MG/Maverick/Puffin, and that the follow-up list is accurate.