Edit: turns out that currentcolor it not the main problem, since most test fail even with direct colour used as function parameters, with one new extra glitch. The culprit seems to be in combination of currentcolor in rgb() in animation for any property. Attaching static MRE (live) with sequence of staggered paused animated elements to show the abrupt step (or incorrect saturation) in the progression where colour was passed through custom property using function that either reveuved colour as literal or currentcolor . That contrasts with the first simple case where the currentcolor is set to the target colour directly without any functions, producing correct "reference" gradual colour progression. This testcase does not even use alpha channel.

Thanks for the report and the expanded repro case in comment#2. There's a couple different things going on there:

• The sudden jump from green to purple with an unresolved color-mix() is a known issue tracked in https://crbug.com/40940960. The sudden jump with an unresolved relative color is basically the same issue.
• The change in interpolation with a resolvable-without-currentcolor color-mix() or relative color appears to be a new issue. I'm guessing that the use of a relative color is changing the interpolation color space we use for some reason.

For the fifth test case in comment #2, I believe it has been fixed in CL https://chromium-review.googlesource.com/c/chromium/src/+/6354303.