Try saving different VFBs. One as a master file. Another for TTFs, a third for OTFs.
I recall Mark Foley telling me that Dalton Maag maintains production VFB sources with TTF (quadratic) outlines in one layer and CFF (cubic) outlines in the other. That seems to me good practice in terms of maintaining a dual format source.
IIUC, reordering the lookups (even if worked) will not give the desired effect since I still want mkmk to be used for all other bases.
Right, so the secondary mark-to-base lookup, after the mkmk, only contains baseA; all the other bases are already processed in an earlier mark-to-base lookup. Mark positioning is iterative, so you can order lookups to move marks in sequence.
The GSUB method you suggest is definitely a viable option for this sort of thing. It's what I do in Hebrew fonts for the furtive patah bases.
I begin every project quizzing the customer about their needs, because generally they have only partially determined these. I try to build as complete a brief as possible, both in design and technical terms, with the focus on documentation of the proposed glyph set. This is a lot of work, typically undertaken speculatively unless the customer has specifically requested my consultation and agreed to pay for this upfront work. I like to prepare in this way so that neither the customer nor I have nasty surprises during the project, and I can therefore price the work both accurately and competitively. I quote design (glyph creation and spacing) on a per glyph basis, making distinction between base, composite, and derived glyphs (derived glyphs would be technically not composites, but based on existing glyphs, e.g. /Hbar/. Other work, such as kerning, OpenType programming, mastering, and testing are quoted based on time estimates.