Persistent audio clicks/pops after every audio burst in Linux and Windows guests on Apple Silicon Macs (UTM 4.7.5 / 5.0.3)

## Describe the issue

Audio recording and playback in multiple guest operating systems exhibit audible clicks/pops immediately after each audio burst, spoken word, or sentence when running under UTM.

The issue is reproducible across:

* Ubuntu 24.04 LTS (ARM64, ubuntu-24.04.4-desktop-arm64.iso)
* Ubuntu 26.04 (ARM64, ubuntu-26.04-desktop-arm64.iso)
* Windows 11 (ARM64, Win11_25H2_English_Arm64_v2.iso)

The click/pop does **not** occur continuously. Instead, it occurs at the end of each audio burst or transition between speech and silence.

Examples:

* Speaking a sentence into Audacity results in a click/pop immediately after the sentence ends.
* Recording with `parecord`, `pw-record`, or Audacity exhibits the same artifact.
* Playing back recorded audio with macOS's QuickTime Player exhibits the same artifact.
* Windows 11 guests show similar behavior with voice recording and playback applications.

The issue has been reproduced using both:

* VirtIO Sound (`virtio-sound-pci`)
* Intel HDA (`ich9-intel-hda` + `hda-duplex`)

The issue has also been reproduced after extensive PipeWire, ALSA, and Windows audio configuration changes, suggesting the root cause may be in the virtual audio path rather than guest configuration.

## Configuration

* UTM Version: 4.7.5 / 5.0.3
* macOS Version: 26.5.1
* Mac Chip: MacBook Pro M5 Max
* Host Audio: Built-in microphone and speakers

## Guest configurations tested

### Linux guests

* Ubuntu 24.04.04 ARM64
* Ubuntu 26.04 ARM64

Audio devices tested:

* VirtIO Sound (`virtio-sound-pci`)
* Intel HDA (`ich9-intel-hda` + `hda-duplex`)
* AC97

Audio stacks tested:

* PipeWire 1.x
* pipewire-pulse
* ALSA
* Audacity
* VLC

### Windows guests

* Windows 11 ARM64

Audio devices tested:

* Intel HDA

## Symptoms

### Linux

Typical PipeWire observations:

Using VirtIO Sound:

```text
pw-record:
QUANT 1024
RATE 48000

pw-play:
QUANT 1024
RATE 48000
```

Clicks/pops occur immediately after nearly every spoken word or sentence.

### Intel HDA

Switching from VirtIO Sound to Intel HDA changes device behavior but does not eliminate the clicks.

Example:

```text
Audio device:
Intel Corporation 82801I (ICH9 Family) HD Audio Controller
```

Clicks remain present.

### Windows 11

Voice recordings and playback exhibit similar end-of-burst clicking behavior.

## Troubleshooting already performed

### PipeWire

Forced fixed clock:

```text
48000 Hz
2048 quantum
single clock
```

Example:

```text
default.clock.rate = 48000
default.clock.quantum = 2048
```

### WirePlumber

Tested:

```text
api.alsa.period-size = 2048
api.alsa.period-num = 6
api.alsa.headroom = 16384
api.alsa.multi-rate = false
```

### ALSA

Forced:

```text
defaults.pcm.rate 48000
defaults.ctl.rate 48000
```

### Additional testing

* Disabled GPU acceleration
* Disabled device suspend
* Disabled rate switching
* Tried 48000 Hz everywhere
* Tried VirtIO Sound
* Tried Intel HDA
* Tried AC97
* Tried PipeWire native clients
* Tried PulseAudio compatibility clients
* Recreated guests from scratch
* Reproduced on 2 Ubuntu LTS releases
* Reproduced on Windows 11 ARM64

None of the above eliminated the clicks.

## Comparison with VMware Fusion

To rule out guest OS, PipeWire, and application-level causes, the same workload was tested under VMware Fusion Pro 26H1 on the same host machine.

Host:

* MacBook Pro M5 Max
* macOS 26.5.1

Guest:

* Ubuntu 24.04 ARM64
* PipeWire
* Audacity
* 48 kHz audio

Results:

### VMware Fusion Pro 26H1

Recording quality is clean.

The attached `Fusion.wav` recording contains normal speech without audible clicks or pops at the end of words, sentences, or audio bursts.

### UTM 4.7.5 / 5.0.3

The attached `UTM.wav` recording contains audible clicks/pops immediately following most spoken words and sentences.

The artifact is especially noticeable during transitions between speech and silence.

### Conclusion

Because the same workload, guest OS, host hardware, microphone, and recording software produce clean recordings under VMware Fusion but exhibit clicks under UTM, this suggests the issue is likely within the UTM/QEMU audio path rather than the guest operating system configuration.

## Expected behavior

Audio recording and playback should be free of clicks/pops when transitioning between silence and audio.

Please compare the attached recordings:

* [UTM.wav](https://github.com/user-attachments/files/28903664/UTM.wav) (clicks/pops present)
* [Fusion.wav](https://github.com/user-attachments/files/28903669/Fusion.wav) (clean recording)

The difference is immediately audible.

## Actual behavior

A click/pop occurs immediately after nearly every audio burst, spoken word, or sentence.

The issue is audible during both recording and playback.

## Reproduction steps

1. Create an Ubuntu 24.04 ARM64 guest.
2. Select either VirtIO Sound or Intel HDA.
3. Boot guest.
4. Record:

```bash
pw-record test.wav
```

or

```bash
parecord test.wav
```

5. Speak several short sentences.
6. Stop recording.
7. Playback:

```bash
pw-play test.wav
```

or

```bash
paplay test.wav
```

8. Observe clicks after each audio burst.

The same behavior is reproducible in Audacity and Windows 11 voice recording applications.

## Request

Could the maintainers confirm:

1. Whether this is a known issue in the SPICE audio backend on Apple Silicon hosts.
2. Whether UTM 4.7.5 and UTM 5.0.3 use the same SPICE audio transport implementation.
3. Whether there are known buffering, silence-detection, suspend/resume, or sample-rate-conversion issues that could explain clicks occurring specifically at audio burst boundaries.
4. Whether there is a recommended configuration for low-latency microphone recording workloads on Apple Silicon hosts.
5. Whether the attached `UTM.wav` and `Fusion.wav` recordings indicate a known SPICE/QEMU audio issue.
6. Whether the issue can be reproduced internally using Ubuntu 24.04 ARM64 and Audacity or PipeWire recording tools.