Actually I'd not call this "debt" but "assets". It's a proofed system with an incredible track record. Please go ahead and show me any freely programmable system that has been in continuous operation for that long.
Think about it that way: It's a system where over time all the "bugs" have been identified, resolved and failure modes mitigated. If you were to replace the system with something new, you'd likely introduce the same old bugs, and new ones, without zero operational gain. Train tracks are rather fixed in their operation; in fact their design is dictated by the train movements. Changes in possible train movements come with changes to the tracks, and those changes are easily reflected by hardwired circuitry.
There would be operational gain. Modern signalling systems can reduce the time between trains to 30s or less, and computer-controlled acceleration and braking is more efficient and smoother for passengers. Also, at some point trains do wear out — i.e significant parts are worn out, like the chassis, that it makes sense to replace the whole train.
I'm just repeating the press statements of London Underground, who have been investing in new trains and signalling systems for the past 10-15 years. (It follows a long period of under-investment.)
New equipment does bring new bugs, which is why they introduce new trains on 'spare' track at night (without passengers), then with passengers on Sunday nights, then Sunday mornings, etc.
We're talking about the track system here, not the trains. For trains it makes of course a lot of sense to use modern locomotion control technology.
However when it comes to the tracks and signaling, there's only very little to be gained to have the actual signalling being computerized on a standard two track railway with the occassional switch track.
The distance between trains, both in time and space, is determined by the blocking of the tracks. There's a hard constraint on the minimum distance between blocks: For any given train running a track, it must be able to come to a full stop within the length of a block. And for safety reasons there must be at least two free blocks between trains (exercise to the reader: why two blocks? Hint: Emergency braking).
Now there is of course the possibility to subdivide a track into microblocks, where a fixed number of microblocks make up a "block" and the "block" borders are dynamically moving along with the trains on a track, so that there's always two blocks of distance between them. But again, this does not require a programmable computer to implement. It can be done using hardwired circuitry (and I'd feel much safer using such a system); just throw a few 74HCT… ICs onto a PCB and mass manufacture that.
There is also organizational debt: http://steveblank.com/2015/05/19/organizational-debt-is-like...
This article seems to be an example infrastructure debt.