Is it rational to avoid certain software features due to possible bugs?

Undo/redo is totally broken on some of modern forums with too rich formatting. Most of them can easily "forget" what was in a long comment you worked on for tens of minutes after clicking a "send" button if some network issue happens (and good luck for those who relies on heavy webpages with hundrends kB of JS code).

For me it is very rational to avoid some of too complicated features in software even if there weren't any bugs have been obserwed before. Especially if software is new and/or proprietary or if I'm working on something important or software has a Turing Mashine in it (Ethereum's Solidity, CSS3, macroses from MS Office etc). Another thing avoided to trust is any software "random" number generator if you have an access to coin or dice for accomplishing your task. Last but not least is not trusting to any connected device at all. So, I do not trust any software if to be honest and I hope my fear might be considered as rational.

We don't allow our admins to use Parted to change partition tables on production servers. We feel that this is rational because the documentation explicitly says that it might be buggy: https://www.gnu.org/software/parted/manual/parted.html "GNU Parted was designed to minimize the chance of data loss. For example, it was designed to avoid data loss during interruptions (like power failure) and performs many safety checks. However, there could be bugs in GNU Parted, so you should back up your important files before running Parted."

I hit Ctrl+S about every two seconds in every editor.

It's a perfectly rational response to all the times we've been bitten by software in the past. (Though I've personally crashed a slot machine, so I may be more cautious than the average bear.)

For a release I'd always do a clean build starting from zero.