I implemented my suggestion: https://hikari.noyu.me/etc/2023-05-24-js-canvas-software-col...

It seems that doing this per-pixel loop on the CPU in JS is slow, but it's not too bad: the first click might take 300ms, subsequent clicks are just 100ms. While this is about ten times slower than I was hoping, it's still fast enough to seem responsive to most users. This is also at 4K resolution, because I wanted a realistic worst-case. It's much faster at 1080p since it has four times less pixels to work with.

You could of course do better with WebGL.