I'm currently using a non-linear version of it, DeepSURV [*], implemented in pycox for a predictive maintenance job. They are much more informative than a binary label and give space to make a business decicision as to how close to EOL you want to take care of the asset.

[*] The underlying neural networks aren't deep at all.

I respect the fact that you can be surprised by that. But even rewinding a bit: where are the data plotted, e.g. time vs failures?