So what are you saying -- that because there are successful sites that require passwords, that passwords aren't a problem for users?
And this problem hasn't killed any sites that were really good? How do you know? Perhaps clever entrepreneurs can steer around the problem, but the point is that there could be a whole other class of apps enabled by a universal identity system.
I happen to work on a site where the user can derive a lot of value from getting recommendations, based on their social network. We do well, but our biggest problem is that a new user doesn't know how good it can be once they've done the work of creating their own network. All over again. For every site that they use.
The short-term answers are strategies such as you mention. Your points about delaying the need for signup are good advice for today.
The long-term answer to this problem is not Facebook, or for everyone to get acquired by Yahoo or Google or whatever. The answer is to create an open or otherwise portable social network -- manage in one place, use everywhere. OpenID, or something like it, will be required for this.
There are some big flaws in OpenID (the phishing problem for one) but at least some people are trying to address them.
I have finally found one clear use case for OpenID: sending out login passwords to newly hired consultants without using cleartext email. Of course, this isn't the kind of killer app that will make OpenID a household word.
Among other things, OpenID is: confusing even to geeks, not especially secure, not solving a particularly important problem, an open-ended set of external dependencies on provider code that you do not control, very poorly branded (because, you know, everyone loves having their ID checked), poorly marketed, and often assumed to be something it isn't.
From a site developer's standpoint, OpenID solves the wrong problem. Nobody is avoiding your site because they have to remember a password. That requirement hasn't killed eBay, Gmail, Amazon, PayPal, Facebook, Plentyoffish, HotOrNot, news.yc, or any other site in history. Frankly, if people are avoiding your site, it's probably because it sucks. Spend your time making your site suck less.
A good way to make your site suck less is to eliminate the need for signups at all, to delay that need as long as possible, or to get the user so excited about the site that they barely notice the signup process. OpenID doesn't do these things. The problem with signups is not the typing, remembering the passwords, or even securing the passwords -- it's the mental overhead associated with deciding whether or not you want another relationship in your life. Can you trust this new site? Will it waste your time? Will it spam you? Will it hit you up for money, forcing you to feel embarassed as you say no over and over? Will your friends find out that you signed up on the site? Will the site plaster your Amazon purchase history all over the open Internet?
Signing up for a site is exactly like making a new friend, and if you think it would be great to automatically be friends with everybody in the world, find a recent multimillion-dollar lottery winner and ask them what that's like. Managing your relationships takes thought, attention, and judgement. Nobody's automated that yet.