For a program that crashes and takes down the entire system as often as Mathematica does it's wild that there's no silent in-the-background autosave tool to keep a temporary duplicate in some scratch directory so if the system goes down you can recover from the crash

@b3m2a1 Years ago in 2015, I had a similar issue and, as strange as it sound, it was produced by a USB accessory that I had (Easy KM switch). After removing it, the crashes when copying and pasting some expressions were resolved. I reported it, but don't have the case number handy.

@gdelfino Unfortunately this issue has persisted over multiple versions and machines for me and is almost certainly related to the front-end struggling to handle complicated copy/pastes, potentially because it failed to fully resolve the box structure of a heavily formatted set of cells. Fundamentally the issue is the fragility of the front-end, which you also see any time you play with complicated FE tricks which are as likely to crash the system as to work

@b3m2a1 if you are able to make a reproducible example, this will go long way towards finding the cause. I do not use the front end as much as I used to. I mainly run long scripts that uses kernel much more than the front end (solving ode's, integration and such). So I do not see the problems you are talking about. I found Mathematica very solid now. I can run it for days and days and never get one crash. In earlier versions this was not the case.

@Nasser the issue is that it's fundamentally not reproducible, if it were I'd have reported it by now and found a workaround. I use Mathematica for complicated derivations and write-ups mostly these days and so the crashes happen largely when copying or cutting expressions or text that have been heavily formatted to make the logic/geometry clearer. On reboot, assuming my work wasn't lost, the expressions can be manipulated without issue

I think the best solution is to just attach a Dynamic task to my notebook that saves every 5 minutes or so. This is likely possible to do without even touching the kernel, so should be fast.