That's a hell of a lot of money to spend 'just in case'. In reality, how often would they be needed - especially in the future when new trains are introduced which will be much more reliable than the 319s. Loops anywhere are of limited use. They're fine if a train breaks down in the right place, but sod all good when it doesn't.
I don't know much about the St. Pancras refit at all, and where the foundations of the station building (Part of the listed structure) lie...But would it have cost them
that much more to cut a slightly wider "box" during construction, assuming it would've been possible?
Of course, anything added to St. Pancras now would have to be a retrofit...And retrofits are time consuming, costly, and highly disruptive. Therefore I doubt that anything significant will be done at St. Pancras for at least the next 75 years, unless it's something bought in as a result of future changes in safety standards, or something else categorising as safety critical.
Are modern trains that reliable though?
<Cough> Class 170's and 458's </Cough>
Can modern trains go into 'limp mode' to get out of a problem? Say a dual redundant setup with the means to allow one or two engines to power, at low speed, to move a train away - even if every other system (except perhaps the brakes!) has failed?
If my understanding is correct, most trains are designed to apply the emergency brakes immediateley if a failure occurs in anything more critical than on-board lighting. Of course such a fail-to-safe option seems good on paper, but IMO it'd be much better for the driver to have the final say-so in what the train does in such a situation.
After all, a train stopping in the middle of the countryside isn't a problem in itself as passengers could be evacuated via the lineside if need be. A train stopping in a tunnel or on a viaduct on the other hand could pose many difficulties - Especially if disabled passengers need to be evacuated as well!

hock:
Personally, I think that trains should have the option to "coast" in the event of a failure, which might allow it to continue to the next available platform where evacuation would be much easier and safer. Furthermore, being able to "coast" a unit would allow failed stock to still run certain services on routes where a rollercoaster style approach to operation could be applied!
