### System details

Omarchy 4.0.0-1 (stable), Hyprland 0.56.2-1, quickshell-git 0.3.0.r20.g28771c7-2, kernel 7.1.8-arch1-3, Intel Core Ultra 5 225H (ASUS NUC15), 2× Apple Studio Display via Thunderbolt + 1× Dell via DP.

### What's wrong?

Returning to a locked machine whose lock screen has blanked (the lock plugin turns displays off via DPMS 5 s after the last activity), waking it with a **single brief tap on a character key** leaves the password field pre-filled with a long run of dots — often dozens. Enter then fails with `Authentication failed`, and you have to clear the field and retype.

### Why it happens

1. The wake keypress is delivered to the lock's password `TextInput` (`misc:key_press_enables_dpms` wakes the displays but Hyprland still passes the key through to the focused surface).
2. The DPMS wake triggers a modeset, which on slow-to-wake displays (Apple Studio Display: several seconds) stalls the compositor, so the key's **release** event reaches the client only after the modeset completes.
3. Wayland key repeat is client-side: the `TextInput` got the press, never saw the release, and types the character at the repeat rate for the whole stall. A 2–4 s stall at ~25 chars/s produces the observed pile of dots.

The dots are all the same character (invisible through masking), so the pre-filled field looks like garbage or even a stale password.

### Steps to reproduce

1. `omarchy system lock`, wait ~6 s for the screen to blank.
2. Tap one letter key once, briefly.
3. When the displays come back, the field contains many dots instead of one.

Deterministic on this machine. Control test: waking with Shift (no character) leaves the field empty, confirming the pile comes from the wake key itself.

Journal timeline of one occurrence (each keypress at the lock triggers `dpms on`, which on this machine always produces a kernel AER burst, making input activity visible in the log):

```
19:18:24 lock-requested (manual lock)
19:18:30 lock blanks — DPMS off, AER burst
         … 46 min, no further bursts ⇒ no input events while blanked …
20:04:49 single wake keypress — DPMS on, AER burst; field already full of dots
20:05:03 unlocked after clearing (Escape) and retyping
```

### Expected

Waking the blanked lock screen with a keypress leaves at most that one character in the field.

### Workaround

Wake with the mouse, Shift, or Escape (Escape also clears the field).

### Suggested fix

Ignore auto-repeat key events in the password field (except Backspace/Delete, where holding to clear is expected). A held key has no legitimate role in typing a password. PR follows.