The thing is that there's huuuuuge road between 'kind-of-working prototype' and 'something deliverable to the client'. And it takes very different kinds of peop

The thing is that there's huuuuuge road between 'kind-of-working prototype' and 'something deliverable to the client'. And it takes very different kinds of people for the portions of it. A "brilliant never-finisher" is great for the idea->'kind-of-working prototype'/PoC, and yes, I think they should struggle to at least get to this, not just throw an idea. But the thing is that this is usually just 20% of the road, and they'd bee useless or even damaging for the rest of the 80%, so it's better to let them move on. And the 'damaging' part is very real (!!): someone who keeps pouring new and better ideas into a project that already has the ideas it needs to be a product is actually bringing negative value by increasing time-to-market, even if everyone only perceives the fact that the product would not even exist without him and that all his ideas are great. By letting the 'brilliant non finisher' start a new project (just give him something challenging and an almost-zero budget until he's really needed again if the company doesn't have the resources for something else), you're avoiding a potential big loss or even a catastrophic failure (imagine the competition brings something inferior to market, grabs to whole market share, and then even poaches the brilliant guys from you afterwards). The only cost to pay would be that you need to keep paying the brilliants for the rest 80% of project duration to keep them for feedback and new projects, while they climb rocks or sip cocktails and you and the rest of the team sweat your guts out (hard to swallow) OR finance some micro-pet-projects to keep their minds working (hard to support).