Yes but the datasheet does not refer to that, and that datasheet has being saying what it says for the last 10 years...
I strongly suspect that whoever wrote the data sheet had simply not realised that it was meaningless to describe a two-terminal optocoupler transistor as operating in common emitter or emitter follower topology, and that the usual differences between those topologies do not apply.
Whether they got the idea that the emitter output connection gives better bandwidth from a simple assumption, based on their misunderstanding, or whether they actually found that the bandwidth of the emitter output configuration is better by testing it, and the difference is due to the lower collector-emitter capacitance because of the higher Vce, I couldn't say.
I suspect your Linear Technology contact is looking for an excuse to justify what was said in the data sheet. He said he's not an expert on optocouplers; I suggest he should have referred the issue to someone who could provide a more definite answer. That said, though, of course he could be right.
and in any case, the "common collector" with the higher vce voltage is surely not going to make "that" much of a difference to bandwidth, compared to the "common emitter" with lower vce?
I don't know how much the collector-base capacitance varies with Vce. I do know that the collector-base junction of the transistor in an optocoupler is deliberately made quite large, because it is the photosensitive area and they want to maximise the CTR. That will increase the collector-base capacitance.
A web search turned up an interesting document at
http://web.mit.edu/klund/www/Dphysics.pdf. At the end of section 1.4 there's a formula for the collector-base capacitance, but there are two variables that greatly affect the dependence on Vce (Vcb actually) whose values, for an optocoupler transistor, are not given, so I can't estimate how much effect Vce (or Vcb) will have on Ccb.