(I saw I got voted down and should explain)
I think the odds of failure at 50-80%, including my own company, and having a bunch of ready ways to re-deploy those people into other projects is good for EVERYONE -- the engineers can move to the most successful projects. If I worked on something for a year or two and it flopped, I'd love to go work at a startup I'd been following the whole time, rather than trying to find a regular job.
I think SFBA in general is basically an incubator/accelerator, if you've been involved in one startup even as an engineer. Formal accelerators help a lot more in places where most of the people who want to do a startup have NOT yet worked for one. College would be the best example.