I think they are just like stripe. When you get to a certain volume you start to do the cost benefit analysis of moving off the saas provider.

Until then they are the first choice for a large portion of the developer population.

After that, it's just a numbers game: will it cost more to redevelop the functionality they provide (either on your own or with another vendor) than you will save over a reasonable timeframe? If the answer is yes, you keep paying twilio, otherwise out migrate.

I have no first hand experience on what size makes sense for twilio, but this article gives us an upper bound (a huge one, to be sure).