When your software is used way after you EOL it.

One of my first jobs was working on a satellite project called ALEXIS at Los Alamos National Laboratory and had been part of a Congressional plan to explore making space missions faster and cheaper. This meant the project was a mix-mash of whatever computer systems were available at the time. Satellite tracking was planned on I think a Macintosh SE, the main uploads and capture were a combination of off the shelf hardware and a Sparc 10. Other analysis was done on spare Digital and SGI Irix systems. It was here I really learned a lot about system administration as each of those systems had their own 'quirks' and ways of doing things.

I worked on this for about a year as a Graduate Research Assistant, and learned a lot about how many projects in science and industrial controls get 'frozen' in place way longer than anyone writing the software expects. This is because at a certain point the device becomes cheaper to keep running than replace or even updating. So when I was watching this USGS video this morning,

I wasn't surprised to see old DEC computers with CRT screens intermixed with newer computers. The LANDSAT 7 was launched in 1999 when DEC no longer existed, but was designed in the early 1990's. The software for running specific hardware on the system was probably written on whatever system (I am guessing an Alpha but I am not sure). As long as that satellite is running, there will be some sort of team working to make sure that hardware has a giant box of spare parts and trying to make sure the software is still running.

Satellites may seem an extreme case, but the same goes for any large scientific studies and many things in the aerospace industry. You can still find inflight TV systems on major plane lines that will reboot themselves to some Red Hat Linux 7 logo.. an OS that was EOL over a decade ago. There are similar items in industrial controllers for making textiles, plastics, and other items.. the devices are large and expensive to replace so will run whatever software was in them for decades. They will also require software which interfaces with them to be 'locked' in place which can have a pile on effect where you find that you need to have some new computer system be able to run something written in Python 1.5.

I expect that a LOT of systems are currently written to work only with Python 2.7 and will be wanting software for it until the late 2030's. The problem is that very few of them are have plans or ability to pay for that maintenance support. While it is very late in the game, I would say that if you are relying on python for such a project, you need to start budgeting your 2020 and future budgets to take in account of paying some group to support those libraries somehow.