[BUG] ACPI EC event workers hang (acpi_ec_event_processor blocked >122s), multi-second freezes on ASUS Zenbook UX3405CA — 7.1.2-3 regression, LTS 6.18.37 fine

### Description

After upgrading `linux-cachyos` 7.0.12-1 → 7.1.2-3, the system intermittently freezes for several seconds at a time (desktop stutters, then recovers). The kernel logs show ACPI Embedded Controller event workers (`kec_query` / `acpi_ec_event_processor`) blocked for more than 122 seconds on an ACPI mutex — multiple workers pile up behind one stuck EC query. During the same window, kwin_wayland reports pageflips arriving ~2.5 s late.

Booting `linux-cachyos-lts` 6.18.37-1 on the same machine: no hangs, no hung-task warnings over a comparable session. 7.0.12 was also fine — the regression came with 7.1.x.

Might be related to other ACPI-flavored 7.1 regressions already reported (#915 TPM boot delay, #911 / #899 input regressions that also work on LTS), but the signature here is specifically the EC event queue deadlocking behind the ACPI global mutex.

### Hardware

- ASUS Zenbook 14 UX3405CA (Intel Core Ultra 9 285H, Meteor Lake-H graphics i915 `7d51`)
- BIOS UX3405CA.314 (2025-11-20) — latest available from ASUS
- CachyOS, KDE Plasma 6.7.2 Wayland

### Versions

| | |
|---|---|
| Broken | `linux-cachyos 7.1.2-3` (`7.1.2-3-cachyos #1 SMP PREEMPT_DYNAMIC Mon, 29 Jun 2026`) |
| Works | `linux-cachyos-lts 6.18.37-1` |
| Works | `linux-cachyos 7.0.12-1` (no such hangs before the 2026-07-04 upgrade) |

### Symptoms

1. Multi-second whole-desktop freezes during normal use (no heavy load, temperatures normal, no memory/swap pressure).
2. Ten `hung_task` warnings in a ~2.5 h session, all identical: `acpi_ec_event_processor` blocked on `acpi_ex_system_wait_mutex`, each "likely last held by" another kworker that is itself stuck in the same path — i.e. the EC query queue is serialized behind one EC transaction that never completes.
3. Correlated `kwin_wayland` pageflip timeouts (i915) during the stall window.
4. On every boot of 7.1.2-3, EC-adjacent init also misbehaves: `upower.service` needs 3 attempts (2 × timeout) and UCSI fails init.

### Logs

Hung task (one of 10 identical traces, all `Workqueue: kec_query acpi_ec_event_processor`):

```
Jul 06 17:12:58 kernel: INFO: task kworker/14:1:14826 blocked for more than 122 seconds.
Jul 06 17:12:58 kernel: Not tainted 7.1.2-3-cachyos #1
Jul 06 17:12:58 kernel: task:kworker/14:1 state:D stack:0 pid:14826 tgid:14826 ppid:2
Jul 06 17:12:58 kernel: Workqueue: kec_query acpi_ec_event_processor
Jul 06 17:12:58 kernel: Call Trace:
Jul 06 17:12:58 kernel:
Jul 06 17:12:58 kernel: __schedule+0x84b/0x2010
Jul 06 17:12:58 kernel: schedule+0x55/0x180
Jul 06 17:12:58 kernel: schedule_timeout+0x31/0x120
Jul 06 17:12:58 kernel: __down_common+0xd8/0x1b0
Jul 06 17:12:58 kernel: down_timeout+0x63/0x70
Jul 06 17:12:58 kernel: acpi_os_wait_semaphore+0xa2/0x170
Jul 06 17:12:58 kernel: acpi_ex_system_wait_mutex+0x81/0x100
Jul 06 17:12:58 kernel: acpi_ds_begin_method_execution+0x20c/0x390
Jul 06 17:12:58 kernel: acpi_ds_call_control_method+0x1a4/0x3c0
Jul 06 17:12:58 kernel: acpi_ps_parse_aml+0x273/0x5a0
Jul 06 17:12:58 kernel: acpi_ps_execute_method+0x1db/0x310
Jul 06 17:12:58 kernel: acpi_ns_evaluate+0x3aa/0x510
Jul 06 17:12:58 kernel: acpi_evaluate_object+0x284/0x420
Jul 06 17:12:58 kernel: acpi_ec_event_processor+0x49/0x130
Jul 06 17:12:58 kernel: process_scheduled_works+0x240/0x5c0
Jul 06 17:12:58 kernel: worker_thread+0x196/0x380
Jul 06 17:12:58 kernel: kthread+0xfb/0x120
Jul 06 17:12:58 kernel: ret_from_fork+0xee/0x2b0
Jul 06 17:12:58 kernel: ret_from_fork_asm+0x1a/0x30
Jul 06 17:12:58 kernel:
```

The blamed holder (`kworker/6:2:17125`) has the exact same stack — also a `kec_query` worker waiting on the mutex.

Visible freeze, same window:

```
Jul 06 17:13:51 kwin_wayland[1946]: Pageflip timed out! This is a bug in the i915 kernel driver
Jul 06 17:13:51 kwin_wayland[1946]: Pageflip arrived after all, 2519ms after the commit
Jul 06 17:21:28 kwin_wayland[1946]: Pageflip timed out! This is a bug in the i915 kernel driver
Jul 06 17:21:28 kwin_wayland[1946]: Pageflip arrived after all, 2490ms after the commit
```

EC-adjacent failures on every 7.1.2-3 boot:

```
Jul 06 19:16:05 systemd[1]: upower.service: Failed with result 'timeout'.
Jul 06 19:16:11 kernel: ucsi_acpi USBC000:00: error -ETIMEDOUT: PPM init failed
Jul 06 19:16:20 systemd[1]: upower.service: Failed with result 'timeout'.
```

### Additional notes

- No thermal throttling, no MCE, no storage errors; RAM 30 GB with no pressure.
- `hung_task` warnings appear **only** on 7.1.2-3 across 30 days of journal history.
- Searched for similar upstream reports; happy to test patched builds or provide `acpidump`, full journal, or bisect between 7.0.12 and 7.1.2 if useful.

Start by reproducing the hangs on linux-cachyos 7.1.2-3 and comparing them with 7.0.12-1 and linux-cachyos-lts 6.18.37-1. Trace the kec_query/acpi_ec_event_processor path around acpi_ex_system_wait_mutex, then bisect between the working and broken versions. Done means the EC workers no longer hang, hung_task warnings disappear, and the reported freezes and EC-adjacent timeouts are resolved.

Written by the indexing model from the issue text.