I've heard of enough misguided "modernisations" (and failures thereof) that I think the "legacy system" was the part that stayed working throughout, and it's the newer stuff added around it that failed. The old stuff may be old but there's a reason it's old... it's outlasted any attempts at replacing it.
I can confirm this. It's usually the Java/Tomcat boxes that front the TPF box that cause this type of huge meltdown. It's almost never the TPF core. It happens, of course, but they've offloaded most of the functionality to more modern technology. So there isn't much code change in the TPF core. And code change is usually what drives outages.
The counteracting force to that will be that they complexity of the surrounding environment increases and becomes more brittle, or is just simply in the way, as constantly increasing demands for new features drifts further from the capabilities of the old system in the middle.
A few years ago, SAS introduced a new status tier. It took 18 months to introduce into the system (Amadeus). The system may be stable, but those kinds of turn around times for a minor customer service change simply isn't feasible. I don't have numbers, but I wouldn't be surprised if one of the reasons upstart airlines such as EasyJet are competitive is that their IT is comparatively modern and can actively support the organisation while IT is more of a millstone around the legacy airlines' necks.
