IntoMobile suggests mobile networks could “crash” or “crumple” under the pressure of FaceTime demand. I think they’ll be OK, mainly for cultural reasons.
IntoMobile paints a picture of long-deprived mobile users jumping on cellular FaceTime at once when available. The more-curious will but I expect cellular FaceTime will not be a killer app, much as video phones have never really taken hold (starting with the AT&T Picturephone in 1964) for various reasons. A popular app, sure, but not enough to take down the network. If everyone were to jump on FaceTime at once, it would be more a quality issue than a matter of the network crashing. Users would be competing to use identical moderately-high uninterrupted bit streams under all propagation conditions.
Some usage of cellular FaceTime using iOS 6 beta software has produced reports of of choppy video, and quality not as good as using FaceTime on Wi-Fi. That’s not a surprise, as mobile broadband connections are usually not as fast as Wi-Fi connections. Users will train themselves to know that when there’s a solid 3G/4G signal with good voice and data, that’s a good time to try FaceTime. (The operator, or Apple, could provide some more explicit hints as to whether conditions are good.) It’s possible for the operator to force a cellular FaceTime call to have better quality in marginal service areas by allocating more resources to that call, but that costs the operator more for the same number of bits, in that maintaining a real-time video connection draws disproportionately-high physical-layer resources (power, code space, time-slices) from other users.