My first point was followed by a smiley :) It wasn't to be taken entirely seriously (though if we have substantially the same ICANN deciding which entities are ready to run a registry, we shouldn't be surprised if we get substantially the same result, ICANN not only doesn't learn from its mistakes, it won't even admit to them). Who gets .org, and how and why, will be the next test in this regard, and probably the last for some years, barring deciding which registry takes over another failing registry (which I think is a distinct possibility).|
As for my second point, I suspect it will never be proved or disproved as we are unlikely to see such a scenario as more open gTLDs under ICANN. If we see any new TLDs, they will need more than a sunrise, or ineffective and largely unpoliced restrictions, as Ben has documented with .biz and .name (though see here regarding NeuLevel apparently cancelling some 7500 registrations). The only nearly foolproof way to ensure that a namespace isn't gamed and filled with drek is to make it restricted and manually review each potential registration. This gets us very close to restricted domains like .museum and .aero, who together have less than 10,000 registrations to date (they'd have to average $10 net profit per name just to return their investment in ICANN's application fee).
Probably few speculators or defensive registrations live there, but that means not much income either, for the registry, or ICANN. Even assuming such TLDs don't fail due to funding problems, they don't make for much of a revenue source for ICANN so creating a few more restricted TLDs probably isn't worth the bother to them.
So I predict if ICANN or a similar successor is in place we'll either see no new gTLDs for quite a few years, or a situation where a registry is given a number of restricted ones, probably farming out the policing to somewhat relevant entities. In some cases those entities might even be willing to operate at a loss (to themselves, not the registry) simply so that their members have a defined and controlled namespace through which other benefits would accrue (this is the model I suggested before the first rollout). In the latter case, my first point still halfways applies. If there is a way to mismanage it, ICANN probably will. -g