Subscriptions do indeed fund perpetual development in a way that one time purchases don't, but the implication that comes along with that is that features are to be added along the way. If you only released maintenance patches, you subscriber satisfaction would dry up really darn fast.
So now we have this model where publishers charge subscriptions so that they can keep their business stable or growing, and subscribers are demanding features be added so that they're getting value for the ever-growing total cost of ownership. And what do you get from that? Feature Bloat.
The subscription model insists that successful products need to continually grow their code base, complexity, and feature set. The idea of stable streamlined applications that do a few jobs really well and otherwise stay out of the way is very hard to sustain in a world of subscriptions.
The alternative -- which was common in the past and remains common among many (not all) game publishers now -- is to temporarily expand your team payroll with talent that produces the product, and then scale it back to warranty the product with necessary maintenance patches while your emphasis turns to growing the market through sales instead of growing the product through features. Later, perhaps, you create another related product or a successor product.
It can and does still work, and it can make for very high quality products that don't become bloated monstrosities. If subscription fatigue is making the news, I'm sure will see a resurgence of this model soon enough.