You can’t know if it helps, you’re never tried what he’s suggesting - it isn’t implemented.
Honestly I’d rather see a slightly smarter solution. The trouble is that the samples seem to be limited by vsync. I’m quite sure most discrete GPUs can draw a cube faster than at 60 fps though. The solution is to decouple the sampling from vsync and only drop to single samples past the minimum, say 4, when the viewport gets below a certain fps threshold, like 20 fps. So in short:
As long as we’re maintaining 60 fps, we draw all the samples per frame the GPU can handle, blowing past the maximum even, if possible
Below 60 fps we maintain a user settable minimum, like 4 by default,
Below 20 fps (or whatever the user sets), we drop down to single samples.