Imagine deploying your bug fix and having to wait two days to find out if it worked!
Because it’s unnecessary.
It’s not a difficult skill.
When folks are in that situation, they tend to adapt quickly to their reality. But that’s not the reality for the vast majority of developers today.
Thankfully.
I spent about 6 months teaching myself how to tie a set of useful knots, and the reality is by now I can't do most of them anymore because day to day it turns out I just never need to tie a Midshipmen's knot (it's super useful when the siruation arises..which is rarely for an IT worker).
It’s just silicone. Who hard could it be?
There is simply no reason to try doing this in your head. You're worse at it than the debugger is. And I say this as someone who does have the skill. It's just not necessary.
It is annoying to find out that your job failed to run or exited immediately due to a typo or other minor mistake.
Of course ML training (and scientific computing) jobs can take weeks or months to complete. Checkpoint and restart features are important because node or other failures are almost inevitable.