Don't know from where this discussion originates, but good to know. Did you use the Rectifier tool?
In my limited experience, thin gaps tend to occur a) when snap-to-vertex tolerances are too small (human misses the vertex), b) when human changes design on the fly (isecalc need to be rerun), and c) when custom placed polygons adjacent to scaled curved primitives are rounded to the desired number of decimals (the vertices in the subfiles are scaled but not rounded).
In this particular case, I would guess that the problematic rects have been created before rounding and they are rotated such that the vertices of adjacent polygons become rounded to a slightly different value. Perhaps the order of events is the key? To mitigate this type of thin gaps, one should first round to desired number of decimals and only then run Rectifier.
For the general case of interfacing with scaled primitives/subfiles (the custom coordinate is rounded to 4 decimal places while the coordinate calculated from the subfile using the 5-decimal places scaling factor is not rounded), the recommended four decimals appears good enough.
In my limited experience, thin gaps tend to occur a) when snap-to-vertex tolerances are too small (human misses the vertex), b) when human changes design on the fly (isecalc need to be rerun), and c) when custom placed polygons adjacent to scaled curved primitives are rounded to the desired number of decimals (the vertices in the subfiles are scaled but not rounded).
In this particular case, I would guess that the problematic rects have been created before rounding and they are rotated such that the vertices of adjacent polygons become rounded to a slightly different value. Perhaps the order of events is the key? To mitigate this type of thin gaps, one should first round to desired number of decimals and only then run Rectifier.
For the general case of interfacing with scaled primitives/subfiles (the custom coordinate is rounded to 4 decimal places while the coordinate calculated from the subfile using the 5-decimal places scaling factor is not rounded), the recommended four decimals appears good enough.