### What happened

Omarchy ships a WirePlumber rule that auto-connects the `a2dp_source` profile on every Bluetooth device. On any machine with a paired phone, that makes WirePlumber acquire an audio transport the phone never offered. BlueZ refuses with `org.bluez.Error.NotAuthorized`, and WirePlumber's Bluetooth node handler then dereferences a nil node. The Lua error aborts the event hook and **the session manager stops linking audio streams for every device**, not just the Bluetooth one.

The result is complete, silent audio loss with nothing in the obvious diagnostic surface pointing at Bluetooth. On my machine audio was dead for five hours before anyone connected it to a phone that had wandered into range at 05:25.

### The shipped config

`/usr/share/omarchy/config/wireplumber/wireplumber.conf.d/bluetooth-a2dp-autoconnect.conf`:

```
## Auto-connect A2DP playback/capture profiles on Bluetooth devices.
## This helps speakers and receivers expose their audio profiles without
## requiring manual PipeWire/WirePlumber recovery.

monitor.bluez.rules = [
  {
    matches = [
      {
        device.name = "~bluez_card.*"
      }
    ]
    actions = {
      update-props = {
        bluez5.auto-connect = [ a2dp_sink a2dp_source ]
      }
    }
  }
]
```

Two things combine here:

1. The match is `~bluez_card.*`, so it applies to every Bluetooth device, not only audio output hardware.
2. The list includes `a2dp_source`, which is the role for a device **sending** audio to this machine.

The stated intent in the comment is "speakers and receivers". Those are `a2dp_sink`. `a2dp_source` only brings in devices like phones, which is the case the comment does not describe.

WirePlumber ships no `bluez5.auto-connect` default of its own, so this behaviour comes entirely from Omarchy.

### What it looks like when it fires

```
wireplumber[1952]: spa.bluez5: Acquire /org/bluez/hci0/dev_XX_XX_XX_XX_XX_XX/fd0 returned error: org.bluez.Error.NotAuthorized
wireplumber[1952]: spa.bluez5: Failure in Bluetooth audio transport /org/bluez/hci0/dev_XX_XX_XX_XX_XX_XX/fd0
wireplumber[1952]: mod.adapter: can't get format: Invalid argument
wireplumber[1952]: wp-node: failed to create node from factory 'adapter'
wireplumber[1952]: wplua: [string "create-node.lua"]:43: attempt to index a nil value (local 'node')
```

From that point, every new stream lands unlinked:

```
$ pactl list sink-inputs
Sink Input #10444
        Sink: 4294967295
        application.name = "Zen"
```

`4294967295` is `(uint32)-1`, so the stream is attached to no sink. Meanwhile every service reports active, the default sink is unmuted at a normal volume, the ALSA mixer is open, and `pactl info` looks entirely healthy. Native PipeWire clients such as the Omarchy bar keep working, which makes the fault look application-specific rather than global.

`systemctl --user restart pipewire-pulse` does not help. Only restarting WirePlumber does.

### The nil dereference is a WirePlumber bug

To be clear about ownership: the crash itself is upstream. `scripts/monitors/bluez/create-node.lua` does

```lua
42        local node = LocalNode("adapter", properties)
43        node:activate(Features.ALL, function (n, e)
```

with no nil check on line 42. I am reporting that to WirePlumber separately.

The Omarchy side is that the shipped default is what reaches for a transport that was never offered, turning a refusal that should be a non-event into a session-wide audio outage. Removing `a2dp_source` avoids the trigger regardless of when upstream fixes the guard.

### There is a second-order cost

Recovering from this means restarting WirePlumber. On Omarchy that runs straight into [quickshell issue 1052](https://github.com/quickshell-mirror/quickshell/issues/1052): the shell begins spinning on the dead PipeWire socket and grows its resident set by about 47 MB per second. Mine reached 51 GB and filled `/run/user/1000` to 100 percent before I caught it, which then broke `Super+Space`. I have filed the recovery-path half of that as a separate issue.

So one phone drifting into Bluetooth range can cost a user their audio, then their shell, then most of their RAM.

### Suggested fix

Drop `a2dp_source` from the shipped list:

```
bluez5.auto-connect = [ a2dp_sink ]
```

That preserves the documented intent, speakers and receivers auto-connecting, and removes the phone case entirely. If auto-connecting phone audio is wanted, it would be better as an opt-in, since acquiring a source transport unprompted is what BlueZ is refusing.

I have applied this locally and Bluetooth speakers still behave as before.

### Steps to reproduce

1. Pair a phone with an Omarchy machine and mark it trusted.
2. Leave the stock `bluetooth-a2dp-autoconnect.conf` in place.
3. Let the phone come into range and connect.
4. Watch `journalctl --user -u wireplumber` for the `NotAuthorized` acquire and the `create-node.lua` nil dereference.
5. Play audio from any PulseAudio client. It produces no sound, and `pactl list sink-inputs` shows `Sink: 4294967295`.

### System details

- Omarchy 4.0.3-1
- Hyprland 0.56.2
- Kernel 7.2.3-arch1-3
- CPU: 12th Gen Intel Core i7-12700K
- GPU: NVIDIA GeForce RTX 3070 Ti (GA104)
- wireplumber 0.5.17-1
- pipewire 1:1.6.8-1
- bluez 5.87-2
- Remote device: phone, `Icon: phone`, advertising Audio Source (0000110a) and Headset AG (00001112)