I think in the next 5 years someone is going to invent an entire stack for long-lived apps. Where every level is as simple as possible and the only updates that are issued are security updates. We need computers to be able to run 30 year old software without having to stress, and the only way I can think of doing that is to minimize feature development and complexity.
With few exceptions, browsers and servers support the same things they did when the web began. Your old server-rendered Perl app might look a bit long in the tooth in 2018, but there's no reason you can't run it exactly the way you describe.
You can of course buy Rails LTS (though from a quick glance the pricing can be very expensive if you have eg a deploy per tenant rather than multitenant). But gems move on from your version, apis change underneath you, etc.
Keeping a moderately complex rails using say 100 gems with okta + google SSO and various other integrations up and running on a permanent basis can easily consume 1/2 a ft engineer.
So if you're in possession of an app that earns say $150k/year, you have to decide if you want to put that half eng year in or if there are more productive ways to spend your time.
This is even more painful if you bet on, for example, a javascript framework that fell out of fashion.
Build something that doesn't cost much to keep going and let it simmer at least 5x longer than you think. And, make copyright dates self-updating. ;) Furthermore, if it's not losing copious amounts of money, pivot, restructure it or sell it to somebody else.
Shutting down usually == disrespecting money.
Or they become slightly successful and abandon it, like they have self-harm tendencies.