Which is a large cost no question. Especially with the size of the codebase you mentioned. What I'm interested in is how, if at all, the transition process is managed. YouTube's transition from Flash might be an example of this. They managed it by creating an opt in sort of beta experience that they tested for a long time before making it the default.

Another thing I'd be curious about is the line at which the opportunity cost of not making the transition surpasses the cost of not developing new Flash features.