The 80 / 20% rule applies once more - why spend 80% of work on the 20% of users that would benefit from progressive enhancement? Unless those 20% of users generate 80% of revenue, which in all likelihood they don't.
I'm a startup guy so of course I understand this, but on the other hand accessibility is one of those things that takes practice and know-how, so if you never do it of course it seems like this giant mountain to climb. However if you make it a priority, it's not 80% of the work, more likely it influences how you do things, so maybe you spend 20% more effort and end up with a little less whiz-bang. There is a cost, but we shouldn't paint it as an insurmountable thing. This is why we have the ADA, so that people with disabilities don't have to go through life as second-class citizens.
You only have to handle the 20% of users as a special case if you have gone and made some technology choices that excludes them. There is nothing inherent about the internet or web that says that needs to be the case.
A rephrasing might ask why we make technology choices that exclude 20% users?
If we're not able to produce a single page web app that reaches 100% of users, that's cool... just make a many page web site without lots of JavaScript that does.
My biggest gripe with this, though it seems sensible on the surface, is that these numbers are often simply pulled out of someone's ass to argue their stance. Often a manager or a front-end person who just wants to work with the new hotness. How do we know 20%(or whatever the number is) is at all a relevant estimation? Who are the demographics that this number makes up? Why is it important to isolate them or exclude them? Furthermore, at best this number often only accounts for the known people that the product is already not serving or poorly serving. Not the people who bail immediately upon realizing that this service isn't for them because they opted against an inclusive strategy.
