I guess it's also because of the technical elegance of using closurees for page actions. It's hard to admit that a technique that feels elegant and expressive fundamentally doesn't scale.

That's part of being a mature developer. We've all stood stubbornly by a piece of inappropriate code that we want to use because it's so clean and elegant, but most of us have gotten past that stage of ego development by the second or third year of school or professional life. It goes without saying that Paul Graham has long since transcended that level.

So, the only answer that fits all of the parameters is that he's trolling us.

Or that it's not his priority - that's the stated reason.

Not buying it. Priority or not, who leaves a forehead-slapper like that in their code for several years?

You're not far off to assume it hijacks your thinking process and steers it to the unimportant.

The change to the code base would be easy. You can increase the expiration timeout or keep around more closures. It would come at the expense of using more computer hardware, which is an elegant technique of scaling: you free your human hardware to work on more important things.