If the calendar was written in C++ or many of it's derivatives then signed integer overflow is undefined behaviour and it could technically choose to do anything it damn well wants (unsigned integers actually do have defined overflow behaviour). Something tells me the runtime of a paper calendar is anything but standard :D
It's not the overflow that's the issue, but a calculation failure that causes the date value to be "0" and thus list the date as January 1, 1970. It's happened to me several times with Pokemon GO.
In this case, it is the overflow as 01111111...11111111+1 in binary becomes 10000000...00000000. (Yes, negative numbers in signed integer format are represented as starting with 1)
But yes, some programs erroneously fall back to 0 in case of a null value, which corresponds to 1970-01-01. That's why so many geotags were recorded for Null Island.
Are we still calling it Y2k38? I think it needs some rebranding or we won't be able recruit devs to put in the overtime to fix it so history can decide it was all fake.
I was part of y2k rememdiation and we did some stuff for the next leap year as well and I had heard about work being done for future things. Im not in it anymore but im wondering if some is already done. Thing about y2k is it was a time when the backbone stuff was never upgraded. Banks using mainframes running cobol programs and such. Now it feels like every company changes their software yearly.