As with all of these GPL-style licences, there's a lot more risk if you're not an all-GPL house, and quite modest risk if you're intending to GPL your code anyway. But the AGPL is different because it can affect not just developers building on the code for redistribution but also those who are just using it internally.
I can understand why they did that, in the context of network-enabled services, but nothing in the AGPL seems to restrict the implications to only that context. As the example above demonstrates, people do use it for other things with potentially profound implications that aren't obvious to anyone who knows about the much more common GPL and LGPL but doesn't realise why AGPL is fundamentally different. For that reason alone, I don't think the AGPL should have used a similar name.
Not sure what a cleaner alternative would be in that case. Perhaps just state that explicitly? "You can use the free version of Opa to build open-source webapps [define appropriately], but need a commercial license if your app is not open-source."
I don't see what's particularly dangerous about the AGPL, though. It's not that hard to comply with: if you build a webapp on some AGPL'd software, you need to AGPL the whole webapp. If you don't want to AGPL the whole webapp, you can just not build it on top of AGPL'd software. Pretty easy to avoid! And in any case the usual alternative to the AGPL is not releasing code at all, or releasing it under a "no commercial use" clause, which certainly doesn't give you more freedom. Compare Gitorious's code (AGPL) to GitHub's code (not released) for example. In both cases they're worried that if they MIT/BSD'd or even GPL'd their code, a rival SaaS could leapfrog them starting from their own code.
[edited the 2nd paragraph above for brevity]