• Measure real hibernate-to-completion time on the actual SBC/storage combination via a controlled, stationary test — the single biggest unknown in the current uncertainty budget (see Considerations).
• Find or create a genuine zero-speed/stationary wifi power-cycle baseline for the router (cold boot, not warm reassociation) to separate router reboot time from true neutral-section dead-time in future outage-duration analysis.
• Steventon/Milton: propose gazetteer update using direction-aware brackets, not the old combined midpoint. As of 2026-08-03 there are two tight same-direction brackets in each approach direction (see Findings, “Updated estimate”): eastbound entry (west-side boundary) 51.6169, -1.2654 (~50m agreement, 2 trips); westbound entry (east-side boundary) 51.6182, -1.2729 (~50m agreement, 2 trips). These supersede the old cross-direction combined midpoint (51.6171, -1.2711, now flagged as not a meaningful physical location). Propose writing both boundary points to location-db/gazetteer next time direct write access is available — current MCP access is read-only pg_query.
• Royal Wootton Bassett now has a tight westbound entry bracket (2 trips, ~40m agreement, 51.5373, -1.9242) — see Findings. Priority gap is an eastbound trip, not a repeat westbound one: no eastbound observation exists at all, and the 2026-08-03 eastbound leg passed through with no detectable drop. Iver/Slough still rests on a single (rejected) trip and remains low priority.
• Apply the same “multiple trips agree” treatment to the other candidates once repeated — Royal Wootton Bassett and Iver/Slough currently rest on a single trip each.
• Move the trigger condition from a fixed distance radius to time-to-section at current speed, once real hibernate-write timing is known (see Considerations, “Fixed radius vs time-to-section at speed”).
• Chipping Sodbury Tunnel reopened 2026-08-03 — diversion ended, direct route resumed. This is the fifth gazetteer entry and has not yet had a data point (2026-07-23 trip terminated near Bristol Parkway before reaching it). Next direct-route trip should check for a wifi-drop signature here.
• Tighten crossing estimates using same-direction brackets, not cross-direction midpoints — done for Steventon/Milton (both directions), still open for RWB. Steventon/Milton now has 2+ tight brackets each direction (see above). RWB has 2 tight westbound (entry) brackets but zero eastbound data — its west-side (eastbound entry) boundary is still unknown. Exit-side (west-boundary) brackets remain looser than entry brackets for both sections (a few hundred metres vs ~50m) — worth investigating whether this is a genuine detection-lag asymmetry or a reconnection-timing artefact.
• New unconfirmed candidate: Maidenhead (not in gwml-neutral-sections gazetteer). Recurring drop right at Maidenhead station on two trips now (2026-07-22, 48s; 2026-08-03, 72s) — see Findings. Train moving at speed both times, so not a station-stop artefact, and the outage extends well past a simple wifi handover. Needs repeating on more trips to confirm as a genuine (possibly undocumented) neutral section rather than noise; if confirmed, propose adding it to the gazetteer as a sixth entry.
✓ 2026-08-04: Steventon/Milton gazetteer row updated to the 2-trip combined estimate 51.6171, -1.2711 (was 51.6212, -1.3093), with provenance in the row's notes. Done via the postgres superuser on gravlax — direct write access to location-db/gazetteer now exists (the owntracks MCP role is still read-only, but sudo -u postgres psql on gravlax gives write). Still open above: Maidenhead (unconfirmed, 2 trips) and the reopened Chipping Sodbury row (awaiting its first data point).