I open a lot of double terminals. The big question for me is how many terminals will my eventual deck support, not what do I lose in the collision.
The most obvious case is chapel/anything. And I do mean just about anything. Let's say I open chapel/cut purse; if they collide I chapel the 3 cards and don't cut purse this shuffle. Let's say I had opened silver instead, well now I have 2 coin to spend - on most boards that means buy nothing and I'm still right where I started. The next shuffle the odds of collision go up, but I may have had time to add a village or a sifter. On the flip side, if I don't collide I have a strong attack that gimps the other guy (he chapels one less copper, he is forced to buy something at 4 or 5 instead of 5 or 6). The other chapel case is where I have a non-terminal (e.g. village), chapel/village let's me increase chapel efficiency by 33% in the case of collision, but also adds no buying power (making it much harder to hit 5) and does not attack the opponent. Given that collision is the odds off bet, I'm almost always going to open terminal/chap if the terminal is strong (I can always chapel it later); I'd be more inclined to take chap/silver if there are strong 2's (e.g. haven, hamlet) or if I'm really worried about T6 collisions.
Now for cards that don't trash it gets a bit more complicated. Here rather than sacrificing a 2 coin buy, collision means sacrificing 2 coin off a likely high total already. In this case even though it is an odds off bet, the drop from 6 to 4 or 5 to 3 is big enough to give one pause, but remember it still is an odds off bet for any non-drawing terminal. So the question becomes how much do you value an atypical gold (or forge or whatever) verses a 4 coin card (like a village). If you are headed engine, you will want that village before the shuffle anyways and the gold is of more limited value than in a BM game.
The bigger concern is not the early collision, but what terminals do you want in the game. Strong terminal 5s make it a very bad idea to open double terminal much of the time. E.g. witch means that you have high odds with 2 other terminals of drawing one dead. Even in an engine, if your big terminals are something like embassy, you just need too many +actions to support lots of terminals. Now if I have some nice TfB going into the engine (e.g. apprentice, forge, upgrade), then I may still load up on early terminals (particularly attacks) and just figure they will be recycled into something better later.
Opening double terminal with much more than +1 draw is pretty bad. Now it is an odds on bet to collide and will only get worse. Further, villages don't decrease the odds nearly as much thanks for the need to draw villages live.
So to recap: trash & terminal is normally fine. You don't lose that much if the terminals collide (you'd trash anyways) and gain little by taking the silver instead. Terminal draw is very bad four opening double terminal. Know how many terminals your deck will eventually be able to support. Grabbing an early cutpurse/watchtower isn't so great if you intend to ghost ship heavily.