S3 is a big part of it. If you're storing lots of data, a "local disk" just isn't enough. Now you're spending your time solving problems Amazon already solved.
Occasional downtime just isn't that big of an issue for many startups, who are still looking for product-market fit.
I see what you're saying, and I failed to illuminate that I'm coming from the perspective of someone who is working with a cluster of riak nodes. Riak is a ring topology much inspired by the Amazon S3 design. I'd never trust anything to just one spinning rust device, sure. I was taking a distributed storage architecture (using whatever open source platform you prefer) for granted. But I recognize that S3 predates many of them.
Occasional downtime just isn't that big of an issue for many startups, who are still looking for product-market fit.