### What's wrong?

`omarchy-hyprland-monitor-clamshell` re-applies the internal panel every 2 seconds while docked, but it can only read a `monitors.lua` rule that is keyed by **connector name**. A rule keyed by `desc:` is invisible to it, so it silently falls back to a different scale and then overwrites the panel about one second after anything — the config, the Display panel, `omarchy hyprland monitor scaling` — sets the correct one.

`desc:` is standard Hyprland monitor syntax (documented on the wiki the shipped `monitors.lua` links to) and is the usual way to survive DP connectors renumbering across hotplug. Hyprland honours such a rule; the watcher does not, and then overrides Hyprland with its own guess.

**Symptom:** the built-in display sits at the wrong scale and will not stay changed. `hyprctl reload` appears to do nothing. Setting the scale by hand holds for one sample and then snaps back.

### Steps to reproduce

On a laptop with an external monitor connected (the poll only runs while docked):

1. In `~/.config/hypr/monitors.lua`, key the internal panel by description instead of connector name, with a numeric catch-all below it:

   ```lua
   hl.monitor({ output = "desc:Samsung Display Corp. ATNA60HR07-0", mode = "2880x1800@120", position = "auto", scale = 2 })
   hl.monitor({ output = "", mode = "preferred", position = "auto", scale = 1 })  -- catch-all
   ```

2. `hyprctl reload`
3. `watch -n1 'hyprctl monitors -j | jq -r ".[]|select(.name==\"eDP-1\").scale"'`

**Expected:** `2` — the rule matches the panel, and `hyprctl monitors` confirms the description matches exactly.

**Actual:** `2` for about one second, then `1` forever. The source of the change makes no difference — the poll re-asserts its own value regardless of whether the scale came from the config, `hyprctl eval`, or the Display panel.

### Root cause

In `bin/omarchy-hyprland-monitor-clamshell` (line numbers from `quattro` HEAD, `7be59e1`):

`monitor_rule_regex()` (L69) builds its matcher from `INTERNAL`, which is the connector name from `omarchy-hyprland-monitor-laptop` (L12) — `eDP-1`:

```bash
printf '^[[:space:]]*hl\\.monitor\\(\\{.*output[[:space:]]*=[[:space:]]*"%s"' "$1"
```

`configured_internal_monitor_value scale` (L83) therefore matches only a literal `output = "eDP-1"` rule. With a `desc:` rule it finds nothing, and `configured_monitor_scale()` (L89) walks its fallback chain:

1. internal rule's `scale` — **not found** (the `desc:` rule is not matched)
2. the catch-all `output = ""` rule's `scale` — **found: 1**

`sync_internal_scale()` (L160) then sees active `2` != desired `1` and calls `enable_internal_output 1` (L151):

```bash
hyprctl eval "hl.monitor({ output = \"eDP-1\", mode = \"preferred\", position = \"auto\", scale = 1 })"
```

`omarchy-hyprland-monitor-watch` runs this via `poll_clamshell_state` every 2 s for as long as an external monitor is active, so the correct scale is overwritten indefinitely.

**#6909 is almost certainly the same defect seen from the other side.** With the *shipped* `config/hypr/monitors.lua`, the catch-all is `scale = omarchy_monitor_scale` where `omarchy_monitor_scale = "auto"`. `"auto"` fails `valid_scale()`, so `read_monitor_scale()` (L121) falls past it to the remembered `internal-monitor-scale` state file and finally to a hardcoded `echo 2` — and the same 2 s poll then forces the internal panel to 2 and overrides every manual change "precisely one second after applying", exactly as reported there. So a stock config hits this too; it does not require a `desc:` rule. #6673 looks related.

The deeper problem is that the fallback is silent. When the watcher cannot parse a rule for the internal panel it does not leave the panel alone — it asserts a scale it guessed, every two seconds, and wins over both the config and the user.

### Suggested fix

Resolve the internal panel's rule the way Hyprland does, rather than by connector name only. Either:

- match a `desc:` rule against the internal output's description (from `hyprctl monitors all -j`) as well as its connector name, or
- ask Hyprland for the effective rule instead of re-parsing `monitors.lua` in shell.

Failing that, `sync_internal_scale` could no-op when no rule for the internal panel was actually found, instead of asserting the catch-all or the hardcoded `2`. Overwriting a scale the user set, every two seconds, with a guess is worse than doing nothing.

### Workaround

Key the internal panel by connector name. eDP numbering is stable across hotplug; only DP connectors renumber:

```lua
hl.monitor({ output = "eDP-1", mode = "2880x1800@120", position = "auto", scale = 2 })
```

### System details

- AMD Ryzen AI 9 465 w/ Radeon 880M (amdgpu), ASUS Zenbook S16 UM5606GA
- Omarchy 4.0.0-1, Hyprland 0.56.2, kernel 7.1.8-arch1-3
- Built-in: eDP-1, 2880x1800@120, `Samsung Display Corp. ATNA60HR07-0`
- External: DP-1, 3440x1440@100 @ 1.6, `AOC CU34V5C`

`bin/omarchy-hyprland-monitor-clamshell` on this machine is byte-identical to the file at `quattro` HEAD (`7be59e1`), so this is not local drift.