| Site | Gas Supply | Gen Maintenance | Cumulative Individual Gen Issues | Total |
|---|
| Site | Gas Supply | Gen Maintenance | Cumulative Individual Gen Issues | Total |
|---|
* Gas Supply — every pod at the site drops to ~0 hashrate at the same time (whole-site outage). Named for the most likely shared cause (fuel, not any one machine) — "Low NG Pressure" is in fact one of the more common reasons logged in gen_down_events, but individual windows have not been cross-checked against live NG pressure telemetry.
* Gen Maintenance — every generator on one specific pod is down/under 25kW at once, while a sibling pod on the same site keeps hashing. Excluded for Stout 4 — its generator wattage telemetry reads implausibly low even while the pod hashes at full capacity, a data-quality issue that needs resolving before this pod's Gen Maintenance figure can be trusted.
* Cumulative Individual Gen Issues — one or more (not all) of a pod's own generators are down/under 25kW while that pod keeps hashing on its remaining generators. The most directly-supported of the three labels, since it isolates a single unit's fault against its own pod's healthy output.
† WW, 90-day view — WW has no data before 2026-07-30, so its 90-day figures are computed over the ~40 days it has actually existed, not padded or diluted by the missing ~50 days. Marked wherever the 90-day preset is showing; not an issue at 7 or 30 days, since both windows fall entirely inside WW's history.
* TX group — Butz, Fortson, Ellyson, and Walker are each a real, independent site with two mining pods sharing one generator group, so there's no sibling pod within any of them to compare against — Gen Maintenance is n/a for all four, same reasoning as a single-pod site. Gas Supply Issue for each of these four is scored against its own hashrate (that site alone dropping to ~0), not against the other three TX sites — they share no gas line or physical infrastructure, so requiring all four down together would almost never fire and would hide real outages. Because none of them has a sibling pod, a Gas Supply Issue event here may actually be a maintenance-style outage we can't tell apart from a real fuel-supply cut — read this column for Butz/Fortson/Ellyson/Walker as "whole site down, cause unconfirmed," not a confirmed gas problem the way it is for GN/Will/Stout/WW. Cumulative Individual Gen Issues is the one figure here computed the same way as everywhere else. Fortson has no generator wattage telemetry at all in this database — its figures rely entirely on logged down/up events from gen_down_events, not the <25kW threshold used elsewhere.
These are assigned by pattern, not by a confirmed root cause per event — each bucket is named for who else was affected at the same moment, not from reading an individual fault code. Treat the labels as a first-pass attribution; before publishing this further, confirm against gen_down_events.reason/category and live NG pressure data.
Site averages are the simple mean across that site's pods. Fleet average is the simple mean across all 20 pods (Gas Supply, Cumulative Individual Gen Issues, Total) or across the 19 pods with a valid Gen Maintenance figure (Stout 4 excluded). None of these are weighted by generator count or miner count.
| Pod | Gas Supply | Gen Maintenance | Cumulative Individual Gen Issues | Total |
|---|
| Pod | Gas Supply | Gen Maintenance | Cumulative Individual Gen Issues | Total |
|---|
Source: InfluxDB MARAGON_Hashrate (15-min mean, per pod) · SQLite gen_telemetry.db (gen_telemetry, gen_down_events) for generator wattage and fault intervals · roster (site → pod) from master_config.json; gen→pod assignment taken from the historical grp tag logged with each telemetry reading, not the current-day config snapshot, so generator swaps mid-window are attributed correctly.
Method: a generator is "down" at a 15-min bin if its last reading within 45 minutes is <25kW, or the bin falls inside a logged down/up interval. Gas Supply = all pods at a site simultaneously ≤50 TH/s. Gen Maintenance = all of one pod's currently-active generators down at once while a sibling pod's hashrate is >50 TH/s (requires ≥2 corroborating generators to guard against one stale/never-closed down-event). Cumulative Individual Gen Issues = sum of each individual generator's own down-hours while that pod's own hashrate stayed >50 TH/s, divided by (period hours × generator count) so it's comparable across pods with different generator counts. The three layers are mutually exclusive by construction — Total is a genuine sum, not double-counted.
Data completeness: generator telemetry begins 2026-06-19. The 90-day preset is therefore ~8 days short of full coverage at its start (excluded as unknown, not counted as up or down); the 7-day and 30-day presets fall entirely inside the fully-covered period. Some pods' full generator complement doesn't stabilize until weeks after 06-19 (e.g. Stout's gens mostly start reporting late July), which affects the 90-day preset more than the shorter ones. WW has no data at all before 2026-07-30 — its 90-day figures are computed only over the ~40 days it has existed (not diluted by padding the missing days as zero), so WW's 90-day numbers reflect a shorter, and notably rougher, early-buildout stretch rather than a full 90 days.
7 / 30 / 90-day figures are all precomputed and swapped in client-side when you change a dropdown — no live recompute happens in your browser. Custom date ranges are not available as a self-service feature: this page has no backend to query on demand (read-only DB access, no running service to call), so a custom range has to be pulled and republished as a new version of this page.
Anchor timestamp for all three presets: 2026-10-05 15:00 UTC. Pulled and computed 2026-10-05.