Each individual mainstream user might have a finite number of needs, but the number of needs of users in aggregate is effectively infinite. I agree that I'd rather use 5-10 great apps than 40-50 marginally useful apps, but why do you think the 5-10 apps you find most useful would be the same 5-10 apps someone else does? No one webapp has to appeal to the entire market to be successful, it just has to appeal to enough users to provide an income to cover its costs.
This reminds me of Joel Spolsky's take on the 80/20 bloatware myth. The idea being that bloated programs like Word are bad because 80% of the users only use 20% of the features. But they don't all use the same 20% is the problem, so you can't just remove 80% of the features without pissing off a sizable group of people. I think the same applies to web apps. Nobody needs every one of them, but the right set of web apps for any given person probably isn't identical to anyone else's.
Also - the cost of servicing that marginal niche is now so low that it's economically feasible to build a solution just for them. To use the kitchen example, most people want gas & electric ranges, but a few might want eco-friendly solar-powered cookware. This product doesn't exist (to my knowledge), but it's because the cost of designing, tooling up, and manufacturing it is above what people are willing to pay. In software, a niche solution could probably be coded up by just a couple programmers. It'd have more economic value than the code they wouldn't produce at their day jobs, so it becomes economically feasible.
Actually, it's probably because no one knows how big that market was. If you produced market evidence that there were 1 million people that would buy a solar powered stove in the next year, there would be product rolling off assembly lines in less than 6 months.
Read Seth Godin's book Permission Marketing if you want to learn about market creation and demand-driven design.
This reminds me of Joel Spolsky's take on the 80/20 bloatware myth. The idea being that bloated programs like Word are bad because 80% of the users only use 20% of the features. But they don't all use the same 20% is the problem, so you can't just remove 80% of the features without pissing off a sizable group of people. I think the same applies to web apps. Nobody needs every one of them, but the right set of web apps for any given person probably isn't identical to anyone else's.