The product was very very very MVP and alpha-like, but came at a finished-product price tag. I have to imagine that this strategy reaulted in a very small and non-representative test cohort of wealthy tech geeks in the Bay Area. Many of them no doubt purchased Glass as an in-crowd signal, with little serious intention of dogfooding the product or developing for it.
Hardware at that pricetag can't just be soft-launched like software. When you charge $1,500 for a device in basically v0.1, you're setting people up for disappointment.