There is a problem with this approach I believe. If all votes expire concomitantly, how does the blockchain reach consensus, even for a short period of time? And there is the possibility of huge anomalies immediately after, before some sort of votes coagulates.
If there is a period of grace to form 'new government', while the previous top is still used, it makes sense, but not sure if it doesn't introduce complications into the blockchain code (true, less than a set expiration time would, probably).
Update: Overall, when all objections and limit-cases have been considered, I believe this is worth testing.
Update 2: Here's another potentially serious drawback: HF will become undesirable by top witnesses, because there is a chance they lose their positions. They might vote against a HF not because it's bad for the blockchain, but because of pure self interest of the witness.
RE: Should each STEEM hard-fork reset witness votes?