That's why mission-critical systems have several sets of floppy disks, and disk-multiplication stations.
> Would you? Really? Or would you recoil in horror at the very idea?
Depends. If the old system is certified and has all error modes defined, while the other new system is a black box with exciting new ways to screw up, I'd go old system ten out of ten times. Which incidentally is why NASA uses ancient chips when they build new robotic drones.
> I worked on a large scale DOS-era software virtualisation project where we moved ~20K users onto a Windows + Citrix platform.
Respectfully: How many lives would you have extinguished had your new system failed? How many failure modes did you encounter during your virtualisation project? How many external systems - which also relied on a very specific way of doing things and would have murdered people if talked to wrongly did you interface with?
No need to answer. We have all had such projects. We know things break before, during, and after the switchover. Only in some environments, systems absolutely cannot break, ever. Aviation is not your average 'let's get us a new mail server' migration project.