Replacing old stuff is a challenge. What is the effort to spin up a whole new widget that does the task, and how do we test it sufficiently to prove it has the same results as the old one? Do you need to run it on sky? If so, can you switch it on/off quickly so we can run the old stuff for science and run the new stuff on an engineering night to prove equivalency? It's difficult to replace but not impossible.

The best outcome you can hope for is some new instrument with better capabilities comes online before the old one dies horribly. For computing systems, we have projects to replace the old stuff with new machines and new OS's and take the time to work out the kinks and bits of tribal knowledge we realize all too late that we lost when so-and-so retired. This is not atypical in the computing industry.

In summary: yeah, things can be hard to replace. There's inertia associated with old systems you have to balance the thought to "string it along, ebay/graymarket a replacement" versus reinventing it.