Well at least the datetime library would be a perfectly reasonable place to handle leap seconds. The place that outputs the timestamp to a human readable format wouldn’t at all be that hard pressed to add a second when printing out the local time with timezone considered. It really wouldn’t add that much complexity.

Could you imagine if anyone was stupid enough to think the right place to handle such a thing was the internal counter of seconds rather than the datetime library though? Ha! That’d be ridiculous right! Imagine all the bugs if someone put something that’s only relevant to local timezone time printing there! It’d cause so many unexpected bugs!

Am I misreading something, or isn't that already in 1,000 years?

You are correct, I made a typo in the grandparent comment - it should have been 1k years. Unfortunately I can no longer edit.

At 10^78 years we could use 512 bit time stamps and be safe from overflow and measure all the way down to Planck time.

Time is so /weird/ that I don't know how computers will ever wrangle it comprehensively.

Just consider the weird but valid scenario in relativity where A can occur before B, B before C, and C before A. I'm not sure how you would wrangle something like that in a data format.

Can that happen in relativity? How?

I’m guessing for different observers. No idea how for a single observer all those could possibly be true

Insertion of a leap second every month.