[DRSDavidSoft]

A verified firmware write can lose its connection during EEPROM restoration. Recovery now accepts that retained safe-state transaction, performs fresh flash readback, then restores settings through the existing guarded workflow. Controller command paths also retain the configured AVRDUDE backend and adapter, avoiding an incompatible auto-selected adapter during recovery.

Validation: targeted stable-path Go tests passed for recovery, configuration propagation, and retained failed-programming state. Live board recovery passed exact flash readback and EEPROM verification and cleared the recovery marker.

WIP for #696: the underlying intermittent strip framing fault remains unresolved. Next action is to capture rejected frames during controlled strip output, remove the temporary bench trace, and validate final firmware. The temporary trace is retained locally in Project/UartProtocol.cpp and is currently installed on the board; it is not a production change. A three-pixel red chase is currently streaming for human acceptance. Seat stress, synchronized playback, and the requested effect-designer/diagnostic integration remain pending. No private logs or device snapshots are included.

Based on #693; retain that waveform/pacing fix when merging.

[DRSDavidSoft]

Continuation checkpoint: e23be3d on wip/strip-framing-trace (GitHub ref verified).

The four existing on-board seat mappings remain left up/down and right up/down. Coordinator and Pealayer report connected; no new handset packet was captured during the bounded learning attempt, and no mappings or saved learning preferences were changed. First-button learning has now been rearmed with single-capture stop, awaiting the human A-button press. No new motion command was sent.

Next action: read the captured complete waveform, confirm the human button label and desired seat action, append one mapping preserving the old handset, read back, then continue B/C/D and physical press/release acceptance. Rainbow/MOSFET and the three catalogued lighting effects remain a subsequent phase; none has physical acceptance from this pass.

The checkpoint also preserves the pre-existing, uncommitted UartProtocol bench trace encountered in the diagnostic worktree. It is unvalidated and has not been built/uploaded during this pass. Its original dirty working copy was preserved. The checkpoint worktree and all seven Pealayer worktrees are clean; no stale worktree metadata was found. Existing recovery source remains a48acbd / #697. Private configuration, snapshots and native UI capture stay local. Validation: live read-only status/catalog and git diff --check; no firmware build or stress test.