This write up by Carl Zimmer basically covers anything I could have said. He links to a number of resources, in particular this pdf, which at a cursory glance looks utterly fascinating and very well done. Figure 1 in that pdf sums it all up, I guess, or to quote Carl:
"Each lineage of venomous animals became deadly on its own, independent of all the others. And yet, in the end, their venoms echo each other... These results show that there are a limited number of ways to kill your victim quickly. No matter what genes you borrow for the evolution of venom, they will end up very similar to other venoms."
Another Zimmer piece specifically points out research purporting to show that snake venom genes are much older than snakes, maybe 200 million years old. That gets around but doesn't quite answer your question. An older review tried to coalesce things (as best as he could in '92), spending time focusing on insectivora and dealing with mammals. The theory is that while venom is an excellent advantage, it requires a significant investment and is often slow-working. In a world with sharp teeth capable of tearing, venom may not be necessary. Mammals, for example, might evolve to not use venom, as it may not be suitable for their high daily energy demands. The pdf linked above briefly touches on the concept of "reverse recruitment," where venom genes may be usefully re-purposed for other biological processes.