Burly Botanicals · Device & Mesh HealthBurly Botanicals · Salud de Dispositivos y Malla

Headline

1 · The fleet, split the way it actually works

Radios and sensors are two different fleets and are never added together. A radio is a mesh device — it carries battery, link quality and travel time, and it is what firmware runs on. A sensor is a probe — it carries substrate or climate values and rides the mesh through a radio. Burly runs 226 radios, 232 sensor heads and 11 Climate One stations. Three numbers, not one.

Radio fleet (mesh transport) — 226

Sensor fleet (probes) — 232

Climate One stations — 11

2 · 14-day data-delivery history — the headline

This is what a 14-day audit is supposed to contain. For every hour of the 336-hour window we ask one question of every device bound to a room and a zone: did your data arrive? The answer is reconstructed from the zone series the device itself feeds — a hole in a zone's substrate series is the probe (and the node carrying it) not delivering; a hole in a zone's climate series is the Climate One not delivering. Same technique that produced 117 weeks of zone-data-delivery history at GTI Danville.

The 3 real delivery faults in 14 days

Zero substrate faults. Zero radio faults. Every fault in the window is a PoE-powered climate station.

Delivery timeline — every group, every hour

Crop-transition gaps — explained, not faults

Which devices could NOT be given a 14-day history, and why

3 · The three climate faults

3a · Veg West Climate One — dead 182 hours and counting

Serial H4210001606, the Climate One in Veg West (room 12408). There is no SPA status field to quote this run — and none is needed, because the telemetry is the harder evidence. Its last data of any kind arrived 2026-07-06T23:00:00Z, and the 14-day series turns that into a history: Veg West's air temperature, RH, CO₂ and VPD all stop dead at 2026-07-06 23:51 UTC — every one of them at the same instant — and never come back. 182 consecutive hours of the 336-hour window with no climate data at all.

3b · The 8-hour facility-wide climate outage on 07-05 — a new finding

At 2026-07-05 04:00 UTC, every single Climate One station at Burly went silent at once — Dry/cure, Flower Eastt A1-A4, Flower Eastt B1-B4, Flower Westt A1-A4, Flower Westt B1-B4, R/D Flower Room, Veg West — and every one of them came back 8 hours later. Through the entire outage, not one battery-powered mesh node missed a bin. Substrate delivery across the facility stayed at 100%.

3c · Flower Westt A1-A4 Climate One — isolated 21-hour dropout

4 · Live status, by room

Straight from the SPA /devices/ endpoint — the authoritative status source. Battery flags red below 30%, link quality flags red below 50. Sensor-head serials are masked; only H-prefix radio and station serials are shown.

Flower Eastt A1-A4 · 85 status-reporting devices all online

Flower Eastt B1-B4 · 101 status-reporting devices all online

Flower Westt B1-B4 · 100 status-reporting devices all online

MOM / PROP · 19 status-reporting devices all online

Veg West · 10 status-reporting devices 2 offline

R/D Flower Room · 18 status-reporting devices all online

Dry/cure · 2 status-reporting devices all online

Flower Westt A1-A4 · 13 status-reporting devices all online

5 · Per-device radio telemetry — and why it is only 24.4 hours

218 of 223 radios returned telemetry across 2026-07-13 13:36 UTC → 2026-07-14 13:59 UTC (2 concurrent workers — the endpoint 429s above that).

The two repeaters — the only radios at Burly with a full signal series

Both are H340-generation repeaters, which are battery-powered — not PoE.

6 · The 51 radios that report but belong to no room

These radios are online, carrying 56 live TEROS-15 substrate probes, and they have no room and no zone assignment in AROYA. They are named after cultivars — — which is the signature of R&D or mother stock that was set up in a hurry and never mapped. The consequence is concrete, and it is not cosmetic:

• Their readings land nowhere. No room chart, no zone series, no dryback, no substrate-EC trajectory.
• AUTOMAIT cannot see them. AUTOMAIT fires on a zone's VWC floor. A probe with no zone can never trigger anything, ever.
• They are the single largest block of devices in this audit that could not be given a 14-day history — not because the data is missing, but because there is no room or zone to attribute it to.
• They still consume battery, mesh airtime and licence.

7 · Devices to retire, and one ghost room

Retire list — silent 60+ days, none in a live grow room

The ghost room

8 · What a delivery gap actually costs the grow team

Burly steers in 2-gal coco, and the whole steering model is built on a four-phase day — P0 activation (lights-on to first shot), P1 ramp-up (small shots back to field capacity and first runoff), P2 maintenance (pulses holding field capacity while runoff manages substrate EC), and P3 dry-down (last shot to next morning — the primary steering lever). Every one of those phase boundaries is read off a continuous series. Lose the series and you lose the ability to see the phase.

• Overnight dryback becomes uncomputable. Dryback is the peak-to-pre-irrigation-low drop, in absolute percentage points of VWC. It needs the peak and the trough. A gap that swallows either end of the P3 dry-down doesn't make the dryback noisy — it makes it unmeasurable, and the day silently drops out of the average.
• P0 goes blind. The P0 read is "did the plant add 1–3 pp of additional dryback after lights-on?" — the single best test of whether the crop is transpiring at all. Worthless without continuous data across lights-on.
• Substrate EC is the lever, and the lever disappears. Substrate EC is a steering output, not a feed setting — it stacks into early flower, pulls back in bulking, restacks in ripening. That trajectory only exists as a continuous series.
• AUTOMAIT stops steering. AUTOMAIT fires on a zone's VWC floor. No zone data, no trigger — and a probe with no zone assignment at all (§6) can never trigger anything.

9 · Method, and what could not be measured

How the 14-day history was built

1. Roster — GET /public_api/devices/?facility=3920&pagination=0 → 469 records. Thin: id, facility, room, zone, serialNumber, modelKey and nothing else. No status. But it carries the load-bearing field — each device's room and zone binding.
2. Zone series — GET /public_api/rooms/{id}/chart/, stitched in 2-day windows across 2026-06-30 → 2026-07-14, for all 10 rooms. This endpoint does retain 14 days.
3. Join — each device is attributed to the zone series its own model feeds. Presence in an hourly bin = at least one non-null point in that hour. Absence = it did not deliver. 336 bins per group, 111 groups.
4. Classification — every gap is classified crop-transition / device fault / facility event, cross-checked against the harvest-group record in the companion cultivation audit.
5. Live status — SPA GET /devices/?facility=3920 (paginated) → 0 status rows.
6. Per-device telemetry — GET /public_api/devices/{id}/chart/?bin=hour for all 223 radios, 2 concurrent workers (429s above that).

Stated plainly: what this report does not know

• No 14-day battery or RSSI history exists for any device, anywhere. The per-device chart endpoint retains ~24.4 hours. Proven three ways in §5. This is an API property, not a pull failure.
• No radio-signal trend for the 221 nodes. They emit only battery_v and travel_time.
• Gateway and sink emit no device-series at all — the endpoint returns an empty payload. Their health is asserted from SPA status only.
• Individual probes inside a multi-probe zone are not separable. The 14-day endpoint is zone-aggregated: where a zone holds four TEROS-15 probes, a hole means all four went silent together. A single probe dropping out inside an otherwise-reporting zone is invisible at this resolution. The delivery group (zone × metric family) is the finest honest unit, and that is what §2 reports.
• 123 roster entries have no 14-day history — 115 of them because they have no room binding at all (§6).
• Product Vault returns no chart series, so the two Climate One chassis parked there can only be seen through SPA status. Rooms with no series: Flower West A1-A4, Product Vault.
• MOM / PROP and R/D bench gaps are inferred as crop/bench moves, not confirmed — those benches have no harvest-group flip date to anchor against. What the data does establish is that their radios never failed.
• Firmware was not audited here. If a WP4 → WP5 push is scheduled: it is pushed over the air by the gateway to every device that is online at the scheduled time, with battery ≥ 30% and signal ≥ 50%. Devices that are offline simply do not receive WP5 — they stay on WP4, unharmed. It is reversible: revert a gateway to WP4 and re-run, with no time limit. The only practical implication for Burly is that the long-dead nodes and dead Climate One stations in §7 will sit the push out, and that is fine.

10 · Actions