Mercerwood bench qualification: 12 active radios, 20 October 2026

21 October follow-up: the user lifted MeshPocket's bootloader-work exclusion. Its fresh Full Companion USB, BLE, contact persistence/sync and three two-way LoRa rounds passed in the MeshPocket qualification. The exclusion and source-bound measurements below describe the earlier 20 October campaign; the other twelve boards were not retested for this update.

Status: PASS. 12/12 active boards have complete accepted bench coverage; 0 have a latest required failure and 0 remain incomplete. The original inventory has thirteen radios: eleven Full Companions and two Full OTA repeaters. The active target is twelve radios: ten Companions and two repeaters. MeshPocket was operator-excluded at this campaign's close for separate bootloader work.

All thirteen 0865 candidate builds passed. Companion updates target 08650f644b680a5dfaacb95142e3cef8529da9d2; the table records each board's actual installed application. Both repeaters retain actual c46defaefd642973cbbdb473815d56b5b261cb2e (C46). Test-only commit 190064f4b5fa97e75dcfca7876568b757ba43b31 passed all twelve CI jobs. Build manifest SHA-256: 1dbf13ae8a3419f2c8e12d0432c37fa990d5bcfbb5649ab4d1d4ed4cec21a468. Seven active nRF52 boards retain the source 88d39510140ce6894cf8e74b93cbd28d7387fa21 bootloader baseline; Companion application-only updates transferred no bootloader or SoftDevice data.

That bootloader baseline describes this campaign's measurements. A subsequent independent deployment restored the original applications after installing an unpublished test bootloader on eight nRF52 boards. Its remaining USB stress and RAK data-preservation limits are recorded in the known-failure follow-up. This does not repeat or extend the source-bound qualifications below.

Board	Installed source	Required USB / BLE / RF / contacts	Overall
Heltec V4.3, FEM off	0865	USB/BLE/contacts PASS; RF retained ABF	PASS
RAK 4631	0865	USB/BLE/RF/contacts PASS	PASS
Seeed XIAO nRF52	0865	USB/BLE/RF/contacts PASS	PASS
Seeed T1000-E	0865	USB/BLE/RF/contacts PASS	PASS
Heltec V3	0865	USB/BLE/RF retained ABF; new contacts PASS	PASS
RAK WisMesh Tag	0865	USB/BLE/RF retained ABF; new contacts PASS	PASS
Heltec T096, FEM on	0865	USB/BLE/RF retained ABF; new contacts PASS	PASS
Heltec MeshPocket	Current state unverified	Historical PASS as of 22:28 UTC only; excluded from active target	Operator-excluded
LilyGo T-Deck	0865	USB/RF/contacts PASS; BLE PASS with WiFi quiet	PASS
Seeed XIAO ESP32-S3	0865	USB/RF/contacts PASS; BLE PASS with WiFi quiet	PASS
SenseCAP Indicator	0865	USB/BLE/RF/contacts PASS	PASS
Heltec MeshTower V2 SD	C46	USB10/3, warm restart and RF3 PASS; SD PASS (read-only metadata)	PASS
RAK 3401 MQTT gateway	C46	USB10/3 retained C46; new warm/RF3 PASS; service PASS	PASS

The safe evidence summary records each axis's actual tested source, artifact, measurement interval, conditions and evidence digest. V3, Tag and T096 retain valid ABF USB/BLE/RF measurements; V4 retains ABF RF. Pocket's earlier ABF and 0865 results appear only in its excluded historical record; its current installed application and bootloader are unverified outside this report for operator bootloader work. The later Pocket tests are recorded separately in the follow-up linked above. ABF is abf7ba5d5c3958580d52d6954d816c08a035a552. For active boards, changed applications, current identities and new 64/32 contact trials are separate measurements. Reviewed source assurance 577aee5162c7553f58bf13b550560da50bbf73860ab37febe3674a5ae73c01aa covers the unchanged behavior; retained results keep their original source, artifact and time. Binary equality and remeasurement are not implied.

Companion USB checks cover 100 query cycles, ten reopens, owned CRUD, persistence through warm restart and applicable radio2 checks. BLE covers three authenticated thirty-second sessions, exact USB-bound identity and device information, clean core state and strict controller lifecycle. Two-way RF verifies exact zero-hop DEV datagrams, counters and owned channel/radio-lease cleanup. Each contact trial verifies 64 owned full-stream records and a 32-record timestamp subset, then removes them and proves absence after reopen while preserving unrelated records. Repeater coverage uses ten query cycles, three reopens, verified warm restart and three nonce-bound RF rounds. Gateway retains its earlier C46 query/reopen baseline; its new held warm restart and RF rounds keep separate intervals and evidence digests. SD coverage is read-only metadata, not media integrity.

T1000-E's eleven physical USB-priority phases passed. These qualify the USB-priority policy, not simultaneous active USB/BLE. The first current cold-host-release strict failure remains recorded before one later explicit unchanged-firmware retest passed. XIAO nRF52's original unaligned QSPI failure was corrected by shared buffer-alignment fix 714cfc4546ac43c220d06abff9adb128f58d16dd; its earlier strict BLE 0x3E failure also remains recorded before the later clean same-bond result.

T-Deck and XIAO ESP32-S3 later passed strict three-session BLE with USB connected and host WiFi quiet; earlier extra-0x3E failures remain in the history. WiFi was checked at trial boundaries, not continuously. T-Deck's initial two WiFi-on failures predate its current boot, so the observations do not isolate WiFi from boot/current-PIN/state differences. Its diagnostic wrapper retained a restoration-check FAIL; later read-only records separately verified handback. XIAO's original wrapper passed handback. Separate exact 64/32 contact trials, secure controller proofs and cleanups later passed with normal WiFi restored. WiFi association at 2437 MHz was observed before and after those trials. No WiFi root cause or firmware fix is established.

V4's earlier contact setup refusal preceded a later bounded-transition trial and exact cleanup. Tower's CAD2 history and Pocket's pre-DFU core8 refusal remain recorded before their separate warm preparations; preparations are not qualification passes. Tower's cumulative receive/read-error count does not prove CRC failure. Gateway's earlier warm failure had 43 complete responsive snapshots with advancing uptime, not a proven hang. Its later CAD2 baseline refusal came after the service was stopped. A separate one-request warm preparation preceded the qualification's own warm restart and RF3; the exact service/configuration owner then returned. Latest accepted results do not relabel those failed attempts.

These finite checks do not establish RF range, phone interoperability, long-duration reliability or every firmware feature. Earlier source-bound results remain in the fleet follow-up, diagnosis follow-up, and initial USB-priority report.