1) Population increase. Where both Reddit and Digg failed was in calcifying around their original idea and being afraid to move beyond it. The Reddit of 2011 is the direct outcome of the Reddit of 2007; Reddit is, at its heart, nothing more than a conversation refiner. When the conversation is held by erudite people interested in talking, what you get is a marvelous idea incubator. When the conversation is teenagers who like fart jokes, what you get is /r/f7u12.

The key, I think, is in recognizing when a change will be necessary before it is necessary and wargaming the possibilities before rolling it out. This is something Reddit did exactly once - with the search functionality. I think if you keep an eye on the next "factor of ten" you'll have a pretty good idea of where you're going. Plan on Hubski with 10x the userbase and you'll have time to worry about Hubski with 100x the userbase. Can you currently handle 10x the users? Then you're good for now. Can you handle 100x the users? No? Then best figure out how you're going to do it.

2) Profit model. Something that Reddit never got on board with was recognizing that if you truly do want a community, if you truly do want to attract content, then you can't just be a news "aggregator." You also have to be a marketplace. Yes, attracting ads is a good thing. As Reddit and Digg abundantly illustrate, indigenous products are welcomed and celebrated... until that person "sells out" (a problem of the power disparity - I could go on about this for quite a while). If you build it into the hub from the get-go, however, you can't really "sell out." Would TheOatmeal do well on Hubski? I think so. Hyperboleandahalf? Certainly. So long as you had some way to quell the "you've sold out" impulse that seems to happen with all anonymous aggregators. Hubski, however, being part tumblr, could easily adopt a DeviantArt model: allow your contributors the ability to sell their wares directly from their hub, for example. Or follow the "self-serve advertising" model Reddit tries at but fails. You could even give it community weighting - the fact that Reddit karma can be redeemed for exactly nothing is remarkably stupid. Allow "power users" to trade their karma for self-serve advertising and (presuming you can build in ways to prevent gaming the system) you suddenly have a powerful model for keeping commerce within your garden.

3) Granularity: I agree with you about symptoms. I think that messing with things, however, is a good way to see how they work. As the site gets bigger, playing around with things like this might be a good way to learn. If you have the ability to build a beta.hubski.com, wherein you sandbox your new content rules on the hubski.com main site, you could easily allow users to evaluate and test proposed changes to the code. This, again, is something neither Reddit nor Digg have done with any real seriousness.

4) Moderation - this is something you should think about more. It's not a problem you have now, but it is likely to be a problem you have in the future.

5) Tags - Current Hubski tags are problematic because there's no taxonomy for them. Also, there's no way to tag something more than once. It makes a "tag cloud" impossible. If I'm following both #politics and #humor, something tagged with both of them should be more visible to me than someone who is only following one of those tags.

Looking at it, I think you stand to benefit from customizability for each individual user, based on coefficients of affinity (#of tags followed, %user is upvoted, etc). I recognize that any time you throw a "coefficient" into the mix you push things completely out of whack... but I'd think pretty deeply about what it might look like if you were to try.