About UR: that 'real time scope' part does not really suffer from the V9958, but more from slightly different timing on turboR.
Oh, I thought that I saw it bad working in a MSX2+ Sanyo Wavy FD (RAM 512KB) in the old days. But, of course, too much time since then has passed. 
That part also "suffers" more with 10MHz in OCM (this seems not to be well "synchronized" or it is too much focused to a 3.58MHz z80).
In software for Mappar RAM, the number of 16KB-RAM is counted.
In the software counted by 8bit register, it fails in the detection of 4MB.
The 4MB has 256blocks.
It overflows when it is 8bit.
The problem is that there is software which won't work and we can't modify that software. For future software I hope it won't happen again. 
Anyway, I think most of actual coders will care of it as you point. We want OCM as a perfect MSX but we have to create compatible MSX software too (it does not happen most of the times though).
As for EMSX_TOP of the latest version, the refactoring of VDP is the main.
The timing of VDP may be getting worse.
However, it is easy to adjust timing.
I am creating the benchmark which measures various timing now.
In the next release, I think that some conspicuous bugs are solvable.
Moreover, in the next release, it returns to 2MB Mapper RAM.
SaebaMSX: Yo can download it for free from here
Thank you!