You're not wrong, it is almost the default to end up with a high maintenance stack and the (perceived) need to churn out new features.
I've done a few hype driven migrations myself.
But if you set out with the primary goal to reduce the amount of time required to run and maintain a software/SaaS/paas/you name it business you can make huge wins in the time department.
Maybe I'm just lucky with a niche subscription based platform that doesn't need much tuning and perhaps 10 support emails a day. It doesn't return a FAANG salary but does net me €90k with minimal operational costs.
The key is not build that app you don't really need (and indeed requires constant baby sitting to please Apple/Google) not rewrite everything to the latest fad eg Svelte + CockroachDB using edge compute on fly.io.
I.e., it's all made-up resume-driven development funded by woo-huffers with too much money that needs to be thrown somewhere.
There are enterprise B2B niches where a single small shop can service hundreds of clients, without needing uber-scale (see: you will never exceed 1mil concurrent users).
You can run everything on a single bare metal server if your dev team knows how to not waste resources (see: no ORM/toy databases, sane/strong-typed and compiled back-end language, use a mature package environment and tools, etc.).
Give a decade, and you can then cash-out when a bigger enterprise B2B shop buys you out. It's not an adderall-fueled manic-depressive roller coaster of a time; but it's honest work.
Best part: the only problem I've ever hit with scaling was people being stupid with their databases.
Even in old cobol codebases coders are still needed to keep up with the changing environment.
The big money in SW is always made in apps and services that are scalable, and that needs constanrt engineering effort.
SW apps that aren't scalable, are a race to the bottom with low margins and low pay that has been offshored or replaced by the services of SW giants.