O-rings are supposed to seal on compression, not expansion.
As it is now, the O-rings are getting blown out of their tracks but still managing to seal the whole assembly quickly enough.
The above unplanned behavior, which is the only thing preventing a hull loss (and a crew loss since there's no provision for escape) is sufficiently iffy that sooner or later we're likely to run out of luck.
(I'd also add about the Columbia loss that NASA had a "can't do" attitude towards the problem they observed of the foam hitting the wing. Hardly a "crew first" attitude.)