Beyond that, if there are a lot of "just barely" off-topic questions being asked, that is a good reason to launch another sub-site; this is how Super User (power computer users, any OS) and Server Fault (sysadmin) sites were launched, because there were a ton of these sorts of questions being asked in the early days of Stack Overflow, but they weren't really about programming.
We've also seen some subsections of the Stack Overflow community want their own sub-site, because they legitimately form a unique, strong community of their own: like DBA (database admins) and Mathematica. All those questions are technically on-topic for SO, but the community wants their own place, with their own reputation system, their own moderators, and their own take on what questions they'll allow.
"The danger of the dilution of the SE network into something like Quora"
This made me laugh, but Quora's problem is that they allow and even solicit pure opinion, whereas on Stack Exchange we prefer answers that can be verified in some tiny way. "Because that's what I think" is a valid reason to answer anything on Quora, which gets into unpleasant weirdnesses like judging answers based on the fame/success of the answerer, because, well, what other science can you apply to an opinion?
Is it possible that most closed questions are "too subjective", but that most unfairly closed questions are "off-topic"?
As someone who's had a few questions shunted from SE site to SE site, being closed at every step, I think SE would do well to emphasize migration over closure. Having a question closed is a real downer, the opposite of all the SE game mechanics. And it's one of the oldest principles of UX: Don't tell me what I should have done. Just do it for me.
These are good points, and as the creator of SE you've undoubtedly put a lot of thought into this. I just wonder, from the perspective of an occasional answerer and moderator, whether there's some way to maximize the value of the questions living in the categorical interstices. Is a question less likely to get the exposure that will lead to a good answer when it's migrated to a smaller sub-site?
And some of the "subjective" questions produce fascinating discussions and enlightening answers; arguably the community moderation of such questions and answers (closing and deletion notwithstanding) still provides valuable weighting that could influence the overall graph and increase the value of the network. Could there be a shadow network still moderated for quality? Hard problems but fun to consider.
Beyond that, if there are a lot of "just barely" off-topic questions being asked, that is a good reason to launch another sub-site; this is how Super User (power computer users, any OS) and Server Fault (sysadmin) sites were launched, because there were a ton of these sorts of questions being asked in the early days of Stack Overflow, but they weren't really about programming.
We've also seen some subsections of the Stack Overflow community want their own sub-site, because they legitimately form a unique, strong community of their own: like DBA (database admins) and Mathematica. All those questions are technically on-topic for SO, but the community wants their own place, with their own reputation system, their own moderators, and their own take on what questions they'll allow.
"The danger of the dilution of the SE network into something like Quora"
This made me laugh, but Quora's problem is that they allow and even solicit pure opinion, whereas on Stack Exchange we prefer answers that can be verified in some tiny way. "Because that's what I think" is a valid reason to answer anything on Quora, which gets into unpleasant weirdnesses like judging answers based on the fame/success of the answerer, because, well, what other science can you apply to an opinion?