I've finished tweaking it, and it's a landsliding success! I got absolutely an smooth video, either 300 fps, or 60 (depending on vsync), with rendering thread able to generate over 2000 frames per second (of a trivial 'fill screen' code, but nevertheless, no stuttering or memory leaks). The 'library' is just 48 lines, and it can be paralleled with no issues to many CPUs. I found that with Arc I don't need to do anything fancy or tricky at all, just [#derive(clone)] and Arc is doing all the magic.
I can't express my gratitude well enough. I've spend about a year trying to find a perfect drawing environment, and I finally have it. Just Arc<[AtomicU32]>.
Thank you a lot!