fix(solis): preserve queued window edits during startup

[mgazza]

Queued time-window edits in #4875 are applied before the first register decode, which replaces the local window cache and silently loses the edit. Drain events after polling and initial decoding, before writing the windows to the inverter.

The regressions exercise the real run() drain, including callbacks queued on a different event loop and continuation after a handler fails. The startup test fails on the original PR and passes with this fix.

Validation: unit_test.py --test solis and all pre-commit hooks for the changed files passed. The two existing unclosed-session warnings remain.

This follow-up targets fix/solis-callback-loop-affinity so its diff contains only the review fix for #4875.

[Copilot]

🟢 Approval recommended

The change is narrowly scoped, matches the stated failure mode, and the updated tests exercise the production drain path and cover the regression case.

Pull request overview

This PR fixes a Solis startup sequencing bug where queued time-window edits could be applied too early and then silently overwritten when the first register decode repopulates the local window cache. It moves the queued-event drain to run after initial polling/decoding but before writing time windows back to the inverter, ensuring edits persist and are applied reliably on the Solis component loop.

Changes:

• Move queued-event draining in SolisAPI.run() to occur after the initial time-window decode and before the control/write phase.
• Update Solis tests to exercise the real run()-based drain path (including cross-loop dispatch) and add a regression that verifies startup decode does not overwrite a queued edit.

File summaries

File	Description
apps/predbat/solis.py	Drains queued entity events after initial polling/decoding so queued window edits are preserved before writes.
apps/predbat/tests/test_solis.py	Updates queue-drain tests to use production run() draining and adds a regression covering the startup decode overwrite scenario.

Review details

• Files reviewed: 2/2 changed files
• Comments generated: 0
• Review effort level: Lite

---

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