Kernel 5.19.12 has never been shipped by Devuan (or Debian): unstable went from 5.19.11 to 6.0.2.

Thank you. You're completely correct.

https://metadata.ftp-master.debian.org/ … _changelog

5.19 is dangerous. It can damage intel laptop display. To avoid.

Yes. Version 5.19.12 to be exact.

Correction: it seems it still isn't fixed in 5.19, at least if I'd had the system running for a while/been hibernated. So have reverted to kernel 5:15 again.

Remember I'm on Chimaera. I only upgraded to backports in the first instance to handle my newish hardware (AMD Zen3 5600G, B550 motherboard) and some issues with suspend/hibernate. If 5.15 works why upgrade further to a newer kernel if that has a regression?

But I an happy to try and pinpoint what that regression is, so that when I do move to a stable Daedalus and no doubt even newer kernel it has been solved.

Not keen on using appimage.

I did try HOS's suggestion to disable hardware acceleration. That didn't solve the problem.

What does seem to have solved the problem is downgrading my kernel to 5:15 from 5:18 .

Personally, I prefer to unpack appimage than to change kernel for one application.

But it's just personal chose, of cause.

I did try HOS's suggestion to disable hardware acceleration. That didn't solve the problem.

What does seem to have solved the problem is downgrading my kernel to 5:15 from 5:18 (I needed a backported 5:15+ image as stable 5:10 doesn't address some amdgpu and motherboard chip recognition issues I had).

I can't see any relevant kernel bug reports in 5:16/5:18 or 5:18 so may attempt to at least identify in which kernel the regression occurred and then report it.

I've also found a 'signature' in syslog that occurs while the kernel is glitching:

Sep 19 17:08:35 grendel kernel: [91634.237980] amdgpu 0000:30:00.0: amdgpu: failed to write reg 1a6f4 wait reg 1a706
Sep 19 17:08:56 grendel kernel: [91655.526407] amdgpu 0000:30:00.0: amdgpu: failed to write reg 28b4 wait reg 28c6

This was previously reported in association with a resume bug back in kernel 5:05 that was then solved.

Any suggestions for what could be causing the problem?

Have you tried disabling hardware acceleration for Zoom?

Reference: https://community.zoom.com/t5/Meetings/ … /m-p/58433

EDIT:

@admin: automatic URL tags are broken.

Currently I'm on the latest 5.11.10 version.
Hhowever recently it [Zoom] become extremely slow starting up, particularly if I've not rebooted the system recently. I'm currently running Chimaera on an AMD 5600G with 16GB RAM and the 5.18.0-0.deb11.4-amd64 kernel.

On my set up at least there does seem a problem with the most recent versions of Zoom.
Start-up is painfully slow, with long pauses while it repaints the screen and meanwhile doesn't respond to the keyboard.
I thought the problem was related to the most recent versions using chromium libraries to write to the screen (signal-desktop uses a similar set-up). I downgraded to an earlier version (5.11.0, downloaded in June) that uses Qt libraries. I thought initially the downgrade sorted the problem but it hasn't. And I'm also now seeing the same problem with signal-desktop.
Hibernating and reawakening (as well the previously mentioned rebooting) does, at least temporarily, restore normal service.
Both Zoom and Signal install in /opt with their own set of libraries. Signal is from the Debian/Devuan repositories.

Any suggestions for what could be causing the problem?

Ogis1975, I never thought about Flatpak.  Thanks for the reminder.