[addaon]

A modern RTC should draw such low power from the battery (tens of nanoamps to maybe a microamp) that the life of a primary battery will basically match the shelf life. Indeed, many coin cells actually have a longer life with a small amount of current draw than in a fully open circuit (in a wrapper on a shelf).

[fluoridation]

How modern are we talking about? This was a Haswell system, I believe.

All I know is that in 20+ years of maintaining computers, hers was the only one I ever replaced a battery for, and since she's left it plugged in 24-7 I haven't replaced it again.

[addaon]

I can't speak for < 2010 or so, but even then 500 nA was common, and half that wasn't rare. I'd expect that we crossed 1 µA around 2003 - 2005, as a very rough guess. But that doesn't mean that every implementation is good, of course; totally possible for a board to do something stupid like have pull-up resistor sinking current continuously on that rail.