The keyword here is "limited" time, ideally more limited than it is now for inventions that don't take off fast enough. Maybe the original inventor doesn't have the skills to perfect it and is just sitting on the patent.
In theory licensing patents solves this problem. If the original inventor doesn’t license or sell the product then owning the patent is literally worthless. Unfortunately, that requires a skilled patent office to reject trivial patents, but not minor but ingenious improvements. Which is in practice an unrealistic standard.
However, I think the balance generally works in the other direction. 20 years feels excessive, but it’s arguably fast enough that taking something from a lab to mass production often eats up 1/2 the time. Further, many patents that are obvious like asteroid mining are going to expire long before their practical. On the other hand simply waiting out the clock to manufacture something means being really behind, thus pushing licensing.
> If the original inventor doesn’t license or sell the product then owning the patent is literally worthless.
That's only true for the public.
A company may not with to license patent use under truly fair conditions, because it would compete with their products using a different technology, they may not with to license to potential competitors, and/or small businesses (fairly).
Alternatively, holder may be using the patent, but choose that their current cash-cow is good-enough so why support competition and innovation, when they can just bump it up slightly every 10 years or so.
Defensive and overly broad patents are definitely a thing, but the intent is to give a monopoly for a limited time. Locking down alternatives not currently in production that are obvious workarounds is kind of a requirement for most patents to be meaningful. The issue is when a company in effect owns an industry and then keeps submitting obvious patents. But, here the problem is with obvious patents being granted.
