Heh, interesting. Perhaps that problem would self-correct after a while; if the payoff for a patent was lower, fewer would apply and there would be less of a backlog... Hmm, looking into it, https://www.uspto.gov/dashboard/patents/pendency.html measures a bunch of different things, but it seems "traditional total pendency" is the main metric, and it seems to be about 2 years on that at the moment. Of course, if there's high variance, then that could mean that a bunch do take 5 years.

That said, if a patent is pending and no one knows when it'll be granted, is that not enough? Would competitors really introduce their own patent-violating version, knowing they may have to suddenly cut production if the patent goes through? Do they do that today?

At any rate, those delays seem absurd; presumably they're due to having a backlog, due to the USPTO not employing enough people, due presumably to them not having enough of a budget. Not that I'm in favor of giving money to patent-granters, though. I wonder if a fee paid by the patent-holders—say, a small percentage of the revenue from selling patented products—would be a reasonable way to fund it.