futurerate.py: requests.get() has no timeout - a stalled Nord Pool connection freezes the main loop indefinitely (follow-up to #1469) #4525

Describe the bug

The Nord Pool day-ahead fetch in futurerate.py (download_futurerate_data_func, currently line 415) calls:

r = requests.get(url)

with no timeout argument. If the TCP connection stalls silently (established but no response - e.g. a mid-transfer network drop, rather than a clean connection failure), no exception is raised, so the surrounding try/except from #1469 never fires, and the call blocks the main prediction loop forever.

This happened to me on 2026-08-15 at midnight (v8.48.1): the update run reached the futurerate fetch and the main loop then hung for over 4 hours until I restarted the add-on manually. Because background threads (Solcast, Axle, history pruning) keep running, the add-on looks healthy - predbat.status stays on its last value with error=false, and no new plan runs happen. Meanwhile the inverter stayed latched in the last commanded state (a force-export slot in my case), which drained the battery to 0% overnight and missed the whole cheap import window.

Log extract (everything from the main loop simply stops after the fetch line; the LoadML timeout 2 hours later is a downstream symptom - it was waiting for the prediction cycle to complete):

2026-08-15 00:00:02.305444: FutureRate: Detected import tariff E-1R-INTELLI-FIX-12M-26-06-13-E agile False export tariff E-1R-AGILE-OUTGOING-19-05-13-E agile True
2026-08-15 00:00:02.305478: Fetching futurerate data from https://dataportal-api.nordpoolgroup.com/api/DayAheadPrices?date=DATE&market=N2EX_DayAhead&deliveryArea=UK&currency=GBP
  [no further main-loop output - no more "PredBat - update at" headers or plan calculations]
2026-08-15 02:06:29.719588: Error: LoadMLComponent: run() exceeded 7200s timeout:
2026-08-15 02:06:29.719688: Warn: LoadMLComponent: run() returned False

This looks like the same failure family as #1469 ("Internet connection failure not trapped, predbat freezes", Sept 2024). That fix added the try/except, which handles connection errors - but a silent stall raises nothing to catch. The bare requests.get(url) has been unchanged since futurerate.py was created in the #1655 code split (Nov 2024) and is still on current main.

Expected behaviour

A hung Nord Pool connection should time out after a bounded period, get caught by the existing exception handler, retry (the existing 3-attempt loop), and fall back to cached rates - i.e. exactly what the #1469 error path already does for clean failures.

Suggested one-line fix:

r = requests.get(url, timeout=30)

(or timeout=(10, 30) for separate connect/read timeouts). I've been running this patch locally on two installs since the incident with no ill effects - a stalled fetch now falls back to cached data as intended.

Predbat version

v8.48.1 (HAOS add-on 1.7.1)

Environment details

• Sigenergy SigenStor EC 6.0 SP, 18 kWh battery, HAOS add-on install
• Octopus Intelligent GO import + Agile Outgoing export; futurerate configured for Nord Pool N2EX
• Incident coincided with the midnight cache rollover, when a fresh fetch for the new date is required

Screenshots

N/A - not a plan issue; log extract above shows the hang.

Log file

Extract above; happy to provide the full predbat.log for the incident window if useful.

Predbat debug yaml file

N/A - not a plan calculation issue (the loop never reaches plan calculation while hung).