This is on macOS Catalina (2018 MBP), latest Nightly. Radeon Pro 555X 4 GB + Intel UHD Graphics 630 1536 MB.

I thought this was a regression at first, but I'm seeing similar behavior even with nightlies from 2019-09-01. On the other hand, I did definitely not see the problem this morning when I first started using some nightly (probably yesterday's) on the same machine, running the same OS. The only thing that I can think of that is different is that I have now plugged in a non-Retina external screen.

Somewhat confirmed that it seems to work alright as long as I don't have my external (@1x) screen plugged in.

Actual results from mozregression: 2019-10-06 bad, 2019-09-30 bad, 2019-09-01 bad.

Could this be related to the swizzle settings code? Perhaps due to the GPU being switched when the external screen is plugged in?

Dupe of bug 1558167.

(In reply to Dirkjan Ochtman (:djc) from comment #5)

Actual results from mozregression: 2019-10-06 bad, 2019-09-30 bad, 2019-09-01 bad.

What about builds earlier than July 30?

This may be related to the swizzling code, in a sense that AMD drivers on macOS 10.15 for some reason don't like it and behave in a broken way (we have a similar situation with Intel IvyBridge on macOS specifically). I kicked off a try with swizzling forcefully disabled, so that we can verify:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=32bee3d53517ef6c491f3458c1c54846d000ca67

Comment on attachment 9099659 [details]
Disable swizzling on macOS 10.15 AMD devices

Revision D48591 was moved to bug 1558167. Setting attachment 9099659 [details] to obsolete.

Clearing the needinfo, since this was closed as a duplicate. If it's still useful to try a mozregression from a longer window, let me know in bug 1558167.