My intuition in this area is based on chips having a specification on maximum soldering temperature and duration. I'm not sure to what extent that is cumulative. I gather the vulnerability is the bonding of the gold whisker wires to the pads on the silicon, but you would want to check that.
Apart from the absolute temperature, chips have a recommended heating/cooling cycle, including heating/cooling rates. That suggests that differential expansion is a factor, which would likely be cumulative (more cycles = more likelihood of fatigue and damage).
The above is intuition, not the hard data you want.
I think what you are doing is a great idea (effectively demanufacturing). I'm hoping you can solve the practicalities, which as far as I can see are quality assurance and being able to guarantee a steady supply of components and a price point below new.
Any plans to retape the components so they can be put though a pick-and-place machine, or are you looking more towards manual rework? I can see that there is room for innovation in efficient ways to get components off boards at volume, as most component removal is in the form of manual rework.
