fix(octopus): keep the last planned dispatches when Octopus returns them as null (#5492)

[springfall2008]

Fixes #5492

Summary

Octopus sometimes answers the Intelligent dispatch query with an error for the planned dispatches alone (KT-CT-4340, "Unable to fetch planned dispatches"): HTTP 200, flexPlannedDispatches: null, the rest of the data intact. The query is made with errors ignored, so the null was read as an empty list and the car's plan was dropped until the next poll. With a car_charging_now sensor that showed as Hold for car with no car charging on the plan.

async_get_intelligent_devices() already keeps the last known planned dispatches when the whole dispatch query fails (since v9.3.1). This PR sends a null planned list down the same path:

• The last known planned dispatches for the device are kept.
• Each is dropped once its end time has passed, and the one in progress is re-trimmed to the time left, so nothing outlives its own end. This is the existing fallback's behaviour, unchanged.
• Completed dispatches in the same answer are still taken.
• An empty list still means Octopus has no planned dispatches, and still clears them.

Telling the error from "none planned"

Across four log files from one IOG site (616 dispatch polls), "none planned" was always an empty list (606 polls) and the planned list was null on 10 polls, every one of them carrying the KT-CT-4340 error. So null is the failure and [] is the real answer.

Behaviour change to check

If Octopus ever returns a null planned list without an error to mean "none planned", a dispatch it had withdrawn would now stay until its own end time instead of being dropped at once. That shape did not appear in the logs above.

Testing

• New Test 15 in test_octopus_intelligent_devices.py: a dispatch answer with the planned list null keeps the cached dispatches (ended one dropped, running one re-trimmed, cache not modified in place) and still takes the completed dispatch; an empty list clears them.
• ./run_all --test octopus_intelligent_devices fails without the fix (Expected the last known planned dispatches kept ... got []) and passes with it.
• ./run_pre_commit passes, including the quick suite.

Notes

• The issue text says a failed fetch was indistinguishable from no dispatches; that was only true of this partial failure. The whole-query failure was already handled.
• Adds a row to tools/debug-journal.md for the symptom and the log shape.
• Not addressed: while the plan was dropped, the "charging now outside its plan" model used the 22 kW rate Octopus reports rather than the configured 7.4 kW. With this fix that path is no longer reached on this error.

🤖 Generated with Claude Code

[Copilot]

🟢 Approval recommended

The focused implementation matches the reported API behavior and includes comprehensive regression coverage.

0 open findings

What changed in this PR

Preserves cached Octopus Intelligent planned dispatches when the API returns null for that field.

Changes:

• Treats null planned dispatches as a partial fetch failure while retaining completed dispatches.
• Adds regression coverage distinguishing null from an explicit empty list.

File	Description
apps/​predbat/​octopus.py	Reuses valid cached planned dispatches on partial API failure.
apps/​predbat/​tests/​test_octopus_intelligent_devices.py	Tests null fallback, cache immutability, expiry, trimming, and empty-list clearing.

🧠 Review effort: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.