Are you sure? Why? Is there some reason to think this?
I've never heard anyone say this.
It doesn't seem right, to me.
Varieties can disappear by remelding with the ancestral population. Years ago I saw reports that the Baltimore Oriole population was remelding with the Eastern Oriole. (Whether that particular process continues I have heard nothing further.)
Humans are doing something similar with our "races," the particular word we have for our varieties. We can do this because, like those orioles, we haven't diverged so far that we're incompatible.
Horses and donkeys, on the other hand, can never re-meld, even though there's a limited (and useful to humans) cross-fertility. A horse and a donkey can make mules and jennies but the process tends to dead-end from there. That's a speciation.
Why?
To put it simply, the possibility for change is dynamic. There isn't a single species which is evolving independent of other evolving species. At each point in time, the opportunities for mutation are different. What's more, evolution is driven to adaptive states. Once an adaptive unit (organism, cell, whatever) has reached such a state, the previous state is no longer adaptive. And, as a multitude of other organisms are coevolving, the previous state is forever lost, and a growing landscape of adaptive possibilities keeps driving evolution forward.