I hope they don't fix this problem by making the new timestamp ridiculously long so it can last millions of years or something.

This was the mistake they made with IPv6 and really hindered its adoption. Just adding one byte to IPv4 would have been more than enough.

I for one use uint32 where it makes sense, because it's a good trade-off with packed struct size and besides - by the year 2106 arrives I will (hopefully) be dead and it won't be my problem anymore.

I wonder if the decision behind using signed int32 was similar? "By 2038 I'll be dead and it won't be my problem lol".

Gonna be fun when some dusty controller takes out a water processing plant and no one can locate it.