I like Mav’s idea. But I think if you are worried about nodes which will only begin misbehaving after a certain ammt of time, then asking them to do minor work to be allowed in would not be outside the scope of work the node would have been doing inside a node group. In that case attacker just has to wait it out.
If anyone here was into NYZO, their queue was months long at one point with nodes were tracked for uptime.
Ideally we want to strike a balance between filecoins need to forge terabytes of data and huge upfront costs compared to someone who happens to be quick with their fingers and end up as an elder running on an S5 galaxy phone. I know the initial idea was anyone can contribute and earn safecoin. But how would one share their resources if the network is likely to be full for nodes at any given time.
I dont particuarly like the proof of human stuff, i was doing the challenges for Idena but their centralized server that awards imploded one week (while ago) and everyone got reset. Idena is like glorified captchas that network participants create then others validate biweekly.
If you want to bootstrap by community effort, its going to end up like bitshares DPOS where everyone was advertising their hardware and commitment before launch to settle on the initial 100 validators.
Casinocoin had members who wanted to run a caching node for them sign an NDA. I mean maidsafe development is a foundation (i think?), we could require initial onboarding for nodes to be run through them.
There has been decent hardware proof coins, currently packetcrypt is doing a bandwidth hard mining algo, or burst whose mining proves hdd storage. I think you may just require some benchmarking and settle on hardware minimums. I assume any lower bound would be servicable by an amazon node but we do need to keep out the rifraf who have spotty and slow conmections or old hardware initially or else we may end up with slow sections.
Anyways, no concrete solutions from me, just reminders of some things which were done in the past.
We need all of this info, there is so much available. Thanks for posting.
Interesting ideas, maybe a way to bootstrap up to a decent network size is for folks to sponsor nodes being run by the foundation (or MaidSafe as I’m not sure the foundation would be permitted to do this under Scottish Charity law)? Incentives for sponsors include helping the network, and maybe earning some rewards (a bit like staking).
Then throttle outsider node joins until the network size is deemed enough to survive an attack. I’ve no idea how big that needs to be but the threshold could be built into the code.
That could turn into some sort of handoff from foundation to the longterm owners hardware some period after launch. Could be interesting could be complicated. 
Worth highlighting. Couldn’t agree more.