As I wrote earlier, the gantry / Y axis acceleration matters more than the speed, so you might have more success (or at least different success) by varying the acceleration at normal speeds.
This test pattern may be help pin down the changes:
Scale it to fit the platform and run it as fast as it will go in Line mode with optimizations turned off and power set to mark a sheet of cardboard.
If the acceleration is now 3000 mm/s², cut it down to 100 mm/s² and it’ll take forever to get up to speed, but will likely track perfectly. Increase the acceleration by factors of two until the problems creep back in; if you can live with a lower acceleration, that’s the solution.
Also, make sure the Ruida controller has the proper Step signal polarity (a.k.a. PWM Rising Edge Valid), as described in this discussion:
The symptoms don’t exactly match what you’ve shown, but it’s worth verifying.