RTL8821AU: remove redundant PR #37 trace-derived post-fwdl pokes

[josephnef]

Cleanup follow-up to #42. Drops 113 lines from HalModule::rtl8812au_hal_init's if (CHIP_8821) post-fwdl block — the trace-replay parity originally added in #37, now made redundant by #42's proper 8821-specific init flow.

What's removed

Inside the if (CHIP_8821) branch:

• Hardcoded T2U Plus MAC at 0x0610..0x0615 (k8821Mac static)
• ~13 trace-captured register pokes (0x004c, 0x004e, 0x0040, 0x0208, 0x0520, 0x0670, 0x0a0a, 0x1874-0x187f)
• BB/AGC value overrides (0x0830/0834/8a4/8b0, 0x0c20-0x0c44 cluster, 0x0c50/0c54/0c90/0cb4/0e90)
• Mislabeled "8814 post-fwdl init" RRSR/queue-ctrl/RA-table block — actually inside the 8821 branch, wrong comment header. Writes 0x0440, 0x04bc, 0x04c6, 0x0520, 0x0524, 0x0670, 0x0990-0x09a4.

The double-writes to 0x0520 and 0x0670 (different values per write within the same if (CHIP_8821) block) were the smell — the trace replay had never been fully reasoned through. PR #42's proper init lands the chip in MAC-RX-ready state from any starting condition, making the replay unnecessary.

Linux validation

tests/regress.py --full-matrix --channel 100 on 8814 + 8821 T2U Plus, VM mode. No regressions; 8821 cells preserved within RF variance:

TX → RX	post-#42 (with trace pokes)	this PR
8814 kernel → 8821 kernel	430 ✓	435 ✓
8814 kernel → 8821 devourer	400 ✓	400 ✓
8821 kernel → 8814 kernel	365 ✓	372 ✓
8821 devourer → 8814 kernel	5865 ✓	5933 ✓

8814-TX cells (2, 4) and 8814-RX cells (7, 8) unchanged at 0 — known separate issues.

What this doesn't validate

Android-side hotplug end-to-end. PR #42 was confirmed by @RomanLut to fix hotplug on PixelPilot + hx-esp32cam-fpv with the trace pokes still in place. This PR removes them. Proper init flow alone should be sufficient (that's the whole point of #42 — driving the chip from any starting state without relying on pre-captured pokes that matched one specific state) — but it deserves re-validation.

@RomanLut — when you have a moment, could you re-build PixelPilot / hx-esp32cam-fpv against this branch and confirm hotplug still works on 8821AU? Single replug cycle is enough. If it regresses, the trace pokes were load-bearing for hotplug specifically (unlikely IMO but possible) and we revert.

What this PR doesn't touch

• 8814AU branch — the if (is_8814a) block at the top of the same area still programs the hardcoded MAC and the DEVOURER_OOT_REPLAY=1-gated path; unchanged.
• The 8821-specific init flow added in RTL8821AU: 8821-specific init flow (replaces trace-derived parity) #42 — that's what makes this removal safe; not modified.

Test plan

• Build clean
• CI matrix (all 5 OS/compiler combos)
• --full-matrix no regressions on Linux
• @RomanLut confirms 8821AU hotplug still works on Android against this branch

🤖 Generated with Claude Code

[josephnef]

Closing — superseded by a scope-tighter rebase.

This PR removed 113 lines from rtl8812au_hal_init's if (CHIP_8821) block, which at the time included a mis-labelled "8814 post-fwdl init" sub-block (RRSR / queue-ctrl / RA-table at 0x0990-0x09a4). PR #57 (merged 2026-05-29) later discovered that sub-block was actually 8814-AU-specific init that had been mis-gated under if (CHIP_8821) — instead of deleting it, #57 moved it to a new if (is_8814a) block with byte-reversed u32 literals. Had #43 landed first, that 8814 init would have been lost rather than rescued.

The remaining content (hardcoded T2U Plus MAC + trace-derived register pokes + BB/AGC value overrides) is still a valid cleanup target — those are genuinely 8821-only and redundant after PR #42's proper init flow. Re-opening as a scope-tighter PR against current master that only removes the 8821 sub-blocks.