This is an interesting idea. I hadn’t really considered approaching it this way. I could see many of the keys that are dropped to put people closer to the new upper bound being used by others to get below the new lower bound. Can you provide insight on how it played out when it was moved from .85-1.15 to the current .65-1.35?
Personally I’m more in favor of removing the lower bound incentive for validators with more than 5 keys. One way of doing that, which shouldn’t be difficult to implement, would be to update the command for adding BLS keys to throw an error if it would give the validator 6+ keys and put them below the current lower bound (0.65 * median stake).
I read your comment (Discussion: reducing BLS keys - #39 by ValidatorONE) and think it’s a great idea, but I wonder how difficult the implementation is. My suggestion above would keep people from adding excessive keys, but it wouldn’t incentivize people to drop keys which is covered in your suggestion.
Perhaps we should open one or more separate threads to discuss these ideas in more detail since they’re a bit off-topic here.