The results so far look promising! I'm curious if this technique can be generally applied to any arbitrary compositing method (like something more ad-hoc than screen and multiply) given the proper training set.
"Godot is/is not the new Unity" is a pretty loaded statement, and would mean different things to different folks.
This particular discussion focuses on the performance gap between the two engines looking specifically at how raycasts are implemented, and the larger implications from that analysis.
I think the use of rotation to visualize the relative speed of clocks was a good design choice both functionally and aesthetically.
That being said, as a hobbyist physicist I can't be too immediately sure about the accuracy eg. @ ~2:50, tho it does appear that the cubes further away are rotating faster which is consistent with the concept of "time ticking slower for the thing moving faster."