I made some updates to the VS. Player. I also looked at simulator implementations of some basic mechanisms (accuracy, speed resolution) and found that they are very far from being accurate (even the vsplayer code has wrong approximations and thus should not be used as reference at the moment).
It appears that the research article on Smogon must be extended to cover such mechanisms extensively, similarly to what was already done for damage calculation.
Without this effort I see little hope to have simulators sufficiently faithful to generate valid battle videos from logs (a process that is otherwise very straightforward and has been tested successfully for close to a year).