Well, I can give you an example where the manufacturer didn't care about the risk:

The Stago STA Compact (Max) automated coagulation analyzer.

The first version of this analyzer ran MS-DOS. It worked fine, but it was a bit difficult to use - it didn't have a mouse. There were some keyboard shortcuts, but mostly I had to use keyboard arrows and Enter/Esc to operate it.

Then there was an updated version (Max) which was basically the same analyzer with new brains: different computer inside, dual-core CPU, Windows XP instead of MS-DOS. It is much, much worse than MS-DOS version.

The database can only hold about 4-5 days worth of results. When it gets almost full, and the sample drawer is open, the internal MCUs timeout while waiting for commands from the main CPU, which gets stuck busy displaying the samples window. And there are race conditions everywhere. If I scroll the results window while the analyzer adds/updates results into it, it gets confused and shows the new results on the wrong table rows, corresponding to other patients - yes, it's that bad.

It's obvious they tried to avoid race conditions as much as possible, for example, it can't print internal control results while the analyzer is running samples, it won't open the samples drawer while running the internal control from the reactives drawer, etc. I would prefer the old MS-DOS system any time.