As I indicated, I know I am going against received wisdom. 
I have seen articles say to use flow rate to control xyz dimensions - which makes no sense to me at all.
Resizing model for each axis individually totally makes sense to me.
I only did this to eliminate xyz tolerances from an issue I am having printing the 3 piece snapmaker quick change toolhead on thingiverse - i still need to try reprinting it after having now done all the recommended calibrations.
In terms of xyz tolerances I got to 20.00 +/- 0.01. The excel bellow shows the starting and end values. The only one of concern to me was the X value as .1 is large enough difference to affect that print. And yes one can have a good debate about whether its possible to measure to that accuracy with calipers, micrometers, given the PLA models surface bends while measuring, bumps caused around the edges of the XYZ letters on the cube, etc etc. But the amount of error here is trivial for almost all things i can think of doing. But is is worth checking that’s the case on any give machine IMO.
I have yet to reach my own conclusions. But this is a fun hole 3D printing has dropped me in
and I am still interested to know if its possible to get a 3d printer to print a truly round 2mm hole on a vertical surface without mucking about with triangles cut outs (or supports inside the hole)