Good question. How long a calculator will be usable depends on two things, in my view:
1. The initial quality of construction.
With a few notable exceptions (the Spice units, for example), the quality of design and construction of HP calcs is very good. My first 28S failed, but that's the only failure I can recall (from a new calculator that I bought myself).
There are some unknowables: is the keyboard in a 35s more or less durable than the keyboard in a 1972 HP-35? Does the higher reliability of an LCD (vs. LED display) overcome the inherently flawed nature of a rubber pressure-strip connector?
2. How easy it is to repair when something goes wrong.
This is the kicker. Once you get into the late 80s and onwards, HP calcs are simply not designed to be taken apart. You can still repair them, but it takes mad skilz and not a trivial amount of luck to do so.
I think the answer's probably that the first couple of generations of calcs are built to a quality level not seen any more in this class of device, simply because they cost a boatload of money. The $395 for a 1972 HP-35 equates to close to $2,000 today.
About the only thing that ever goes wrong with a first-generation HP is LED and keyboard failures, and even those are rare, and easy to fix albeit with donor parts from another unit.
3. How heavily the calculator is used.
A current HP-35s that's sitting on a shelf with no batteries would likely fire up with no problem a century from now, whereas the solderless connections in a Spice will virtually certainly have degraded by then, rendering the calculator useless. But a Spice in good working condition might hold up better over years of day to day use. This is another one of those "unknowable" things.
FWIW, I think some of the older desktop calcs would turn out to be the robustness and longevity champs, if only for their exceptional keyboards.