It doesn't matter that $100 or whatever sounds like a lot of money to me, or that $100 can buy a lot of slices or software or pizza for that matter. What does matter is that setting up cacti/nagios/etc takes time. That is time where I could be doing things that actually make money, like fun SEO work or intellectual-stimulating-as-watching-paint-dry-but-sure-as-heck-profitable button A/B testing.
A single A/B test which offered a 3% lift anywhere pays for $100 a month, and for companies or individuals operating above my scale, you have even less of a bar to waltz over. I also typically don't have to maintain those (gif files break remarkably seldom in my experience compared to software), where getting more software running just increases the number of things that can need tweaking or go wrong next month.
Now, you're almost certainly a better sysadmin than I am, so maybe knocking together some cacti/nagios scripts is really, really fast for you. But I'm betting the best use of your time is probably more effective than doing this.
Also, do you really spend all day doing A/B tests or do you have 'marginal' time that you could dedicate to other things? I think we (maybe not you; I'm speaking in general) make the mistake of saying that since we bill at X dollars an hour, that our time is worth that. It isn't: you can't fill all your hours with pay work all the time.
Offtopic, Cloudkick is a very sexy offering, and it's priced for a certain set of customers, which may or may not fit you or I, but ideally, I'd like for them to find a cost structure that scales a little better, perhaps. 10% of the cost of the server might work better for users with less than 10 servers, for example.