This is what my project was all about. It went for another metric than the company's own press releases saying how good they were. And yes, I got a lot of charismatic operators who were all enthused at the startup. Then I just let time do its thing.
I wouldn't really want a ban on free domain names - it's like the heavy handed answer to the underlying problem. Clients just want a stable place to be. It's a good heuristic that the TLD operations last longer, but as my project helped me understand, having a TLD was no magic bullet either. I would have accepted a winner using a free domain if it won its category fair and square.
Put a little more fancy, don't even put a 10 post requirement etc - just empower the clients more. That's why my project specifically ignored host histrionics, and simply started the clock timer. "If you live, I'll notice. If you croak, too bad for you". I think my rules were pretty fair, and I hope some people see it as a useful new concept of service. In other industries it's called "Attestation" - I Attest that my winners survived the notorious burnout cycle of 4-7-9 months, and so are statistically more likely to be good hosts. I provided a little advice to help the hosts, but otherwise I'm neutral - no particular agenda with the entrants. So there they are.
Sure, there are more good hosts that did not enter my project for various reasons, and ones that emerged after I closed it, but of the winners I picked, I basically only had one "bad guess" and that was cWahi's unfortunate data loss incident, but the company itself is still around and they were rock solid up to that point.
I feel I am a pioneer but I will surely not be the only such Attestation service. So then what happens is when a new host shows up, it is sorta chicken and egg, they need clients, but clients need to know that they won't hit that magic zone when they all drop like flies. So very roughly, just check a host's sign on date, and maybe with a little searching make sure it's a year old with the same company name. Basically my project in a nutshell says that if a host lasts a year, they're worth looking at.