This is an interesting idea, but I don't think it'd be practicable, for a few reasons.
First and foremost, on an OOC level, the addition of a risk of death to escape plots came about to balance some serious issues in past versions. Basically, there was no risk to attempting to get out, meaning that there was nothing to discourage handlers from trying again and again. This, in turn led to a ridiculous number of escape plans of varying degrees of possibility, most on the lower end. In my opinion, the current rule should've been around for ages, but unfortunately hadn't really been thought out.
Moving beyond that, the practical issues I see are as follows: it's been established that, to as great an extent as is possible, the terrorists do test each collar individually. This means that the odds of one slipping through while defective are very low. I'd say we're talking possibly even lower than a 1/10,000 chance, given the established efficacy of the collars.
This applies only to things that would actually detonate the collars. I suppose a totally defective collar slipping through would stop the terrorists from detonating it at range as well, but by the same token it would also stop a character from being detonated in a DZ, or if somebody grabbed their collar.
In effect, putting such a situation into practice would mean either a. rolling every time a character got into a situation in which their collar might go off, which would involve slowing the game down a lot (rolls need at least two staff present, and the ideal is three including an admin) or b. pre-rolling the collar functionality of every character in the game, which would be a very awkward situation for staff, who would (on the unlikely offchance someone actually hit a 1/10,000-100,000 chance) have to keep quiet about it.
Basically, as I see it, the payout is very low compared to the work and time needed for it, especially since it largely boils down to the potential for rewarding handlers at random to an even greater extent than normal.
I'm not a fan of modifying odds based on character actions either, because anything that won't turn the collars off full-stop will make them explode. Similarly, taking preparations that tip off the terrorists may result in collars getting popped before any action can even really be taken.
As always, staff will be honoring any attempts that would work within the established constraints of the collars and the game. As I see it, that's a pretty good possibility for reward, better than random odds. We have an established list of things that will not work for escapes (and it is pretty long and extensive!) but I am also very confident that not everything has been considered. I'd rather see any escapes which occur come about through channels of ingenuity rather than luck.
I speak only for myself here, though, and this is not a staff ruling, just my thoughts.