**What happens.** Every time the idle lock fires while the screensaver is up, `ttfx` aborts and systemd-coredump writes a ~1 MB core. 25/25 ttfx core dumps on this machine occurred within milliseconds of `omarchy-system-lock`.

**Expected.** The screensaver shuts down cleanly on lock, as it does on dismissal.

**Cause.** `omarchy-system-lock` ends with `pkill -f '[o]rg.omarchy.screensaver'`, which matches the screensaver *terminal* but not `ttfx` (whose cmdline contains no such string). Killing the terminal closes the pty master; ttfx, rendering at 120 fps, hits EIO on its next write, tries to report it via `eprintln!`, and that write also fails:

```
thread 'main' panicked at library/std/src/io/stdio.rs:1166:9:
failed printing to stderr: Input/output error (os error 5)
```

(Panic text recovered from the core's memory; the binary is stripped so the backtrace itself does not symbolize.)

`omarchy-screensaver`'s own `exit_screensaver()` already uses the correct order — `pkill -x ttfx` first, then the terminal — so the dismissal path is clean and only the lock path is affected.

**Timeline of one instance:**

```
13:14:12.860  idle: lock-system: lock-timeout -> omarchy-system-lock
13:14:13.002  lock script exits 0
13:14:13.008  ttfx aborts (SIGABRT)
13:14:13.043  hyprland: failed to read client connection (terminal)
13:14:13.492  lock secure=true
```

**Repro.** Let the screensaver start, then let the idle lock deadline pass without touching input. `coredumpctl list | grep ttfx`.

**Fix.** Kill `ttfx` before the terminal in `omarchy-system-lock` (mirroring `exit_screensaver`), and/or make ttfx exit quietly on EIO/EPIPE from the output tty.

**System.** Omarchy 4.0.0.r1687.gf5893dd-1, ttfx 0.3.1-1, terminal `foot`. Backtrace is unsymbolized — ttfx ships stripped and is not on debuginfod.archlinux.org.

Filed by Claude Opus 5 via Claude Code.