Random fact: there's 12 packages that don't build for ARM in MeeGo Trunk (fails to compile). We've started patching those and most are trivial build errors. So from a build perspective, we're not much different. On image building side (mic2 support) ARM is a equal member as well now.
Short story: because it gets a strong ARM offering in MeeGo which is useful for many various reasons, such as platform adaptation. Who'd want to adapt a platform where the ARM reference device works like crap? Case: Why are we implementing power saving in our N900 kernel when there's no real benefit to developers in this? Fact: Because we're making a complete hardware adaptation, not a 'developer only' one. A hardware adaptation has to be full to be stable or not be worth anything. Rest of OS, that's another thing.
Agreed, this might be a problem, but let's see how we can solve this in a sane manner. To be honest, it should be entirely possible to keep this going. Even at what, 2 man hours a month to check the lights are still on and things are building and getting exported as redistributable bits at repo.meego.com. Think we're overestimating those few blob's importance. And as I've repeatedly said, I'd like to get rid of them if I could. But as daperl once said: So, we have them to give a proper user experience. I'm hoping to get them to a redistributable license.