One thing that I think they want which current tools aren't so good at is speed.
It seems that the sort of logging that this is used for is debugging traces. i.e. hundreds of messages that you will almost never want to look at - so it needs to almost be free.
You only look when a problem occurs and then you want to see everything that happened in the last few minutes so that you can piece the history together.
So passing messages into the kernel needs to be fast and pulling them out simply should not happen in normal usage - they just get discarded.