I would like to know if there is a prectical limit in number of replica restore point.
The space consumed by the differential disk is an obvious problem, and from what i've seen is more an incremental forever than a reverse incremental (the terminology is obviously translated )
the io from incremental merge is a general problem, no matter how many restore poit i have (and i wonder how refs can help in the matter)
As result of the previous consideration i think that a vm running from a long chain would be much slower...