I got into few troubles with MS FV and while I believe main reasons for them is tool being too old, I'd like to ask them.
1. LTSH table. Spec says that condition for "linear" spacing of the glyph above 50 PPM is that difference between scaled and gridfitted width is less or equal to 2%. It looks that in FV it reads it as "less or equal to 2 pixels" - at least that's when it stopped reporting errors in the table I generate.
2. VDMX table. According to the spec, all glyphs of the font are taken in account when yMax and yMin are calculated. Then gridfitted values are compared to scaled yMax and yMin (the only place for which is in head table). However, it looks that FV checks for some subset of the font, providing very strange values as "correct".
3. gasp table. Basically it needs to be updated to cover version 1. Same is for version 4 of the OS/2 table.