Yeah, people underestimate the value of “finished” software: in an ecosystem with lots of stable dependencies, there’s very little reason for useful software to change constantly.
Even "finished" software needs maintenance. Nothing is ever bug-free so needs fixes. And it doesn't live in a vacuum, the ecosystem evolves and continuous adjustments are needed when APIs evolve or libraries change.
In well-written software, the maintenance burden is low, but it's not zero. Without any maintenance, you can maybe run some piece of software in some closed-off container for a while, but it will keep rotting away and eventually you won't even be able to compile it anymore.
What about "GNU Hello", never finished? Clearly this isn't true for 100% of all software, so the only thing we can conclude is that it either "depends" and/or is very subjective.
> when APIs evolve or libraries change.
If you live/work inside an ecosystem that favor stability over "evolving APIs", you can actually be able to use libraries that are decades old, that doesn't have any bugs for the stuff they expose and things just work. I mostly experience this in the Clojure ecosystem, but I'm sure it's true for other ecosystems too.
Does "small burst of activity and dependency updates twice a year" seem inadequate to you? That's the scale of maintenance that the project in question seems to exhibit, which is what we're apparently calling not maintained.