Some open source startups fail because they are non-scalable service companies deluding themselves that they are product companies. Furthermore, they do not own enough of any durable advantage (IP, talent, mindshare), and so multiple "product" companies pretend to not compete while offering similar services in a niche that does not have that much demand. Furthermore with FOSS, most potential customers are often politically incentivized to poach upstream code and talent, give nothing back and eschew overpriced "consulting" services entirely. So it's almost always the wrong business model, unless you dominate it.
(I've plenty of enterprise FOSS consulting to realize there are easier ways to make much more $/time, like enterprise startups that are product companies. Also, PGs essays about "consultingish.")
Yeah, you are right; look at companies in the hadoop eco system: cloudera, hortonworks, mapr. They are just burning the cash; most of the money comes from the services/consulting. This will succeed if one dominates the sector, the way Redhat does in the area of enterprise linux.
I think this is a big part of it (from someone who has considered a consultancy). These aren't necessarily hugely scalable, which isn't a problem unless you are looking to be a big hit...which is exactly what VCs are aiming for.
(I've plenty of enterprise FOSS consulting to realize there are easier ways to make much more $/time, like enterprise startups that are product companies. Also, PGs essays about "consultingish.")