Often, the cost of switching is too high or too complex to justify. On top of that, many applications commonly run in manufacturing etc., simply does not run on any other OS.
The billions that have been lost, and the lives that have been lost, have, in the blink of an eye, rendered the "too costly to implement" argument moot.
For bean-counting purposes, it's just really convenient that the burden of that cost was transferred onto somebody else, so that the claim can continue to be made that another solution would still be too costly to implement.
Accepting the status-quo that got us here in the first place, under the pseudo-rational argument that there are not realistic alternatives, is simply putting ones head in the sand and careening, full steam ahead, to the next wall waiting for us.
That there might not be an alternative available currently does not mean that a new alternative cannot be actively pursued, and that it is not time for extreem introspection.
