Nothing stood out from the diff in a quick review. I'll be test driving this branch until tomorrow and then do another review.
I think the code looks sane as such, the only problem I fear is some racy oddball edge case that relies on the old startup sequence. That said; I am not overly concerned by the nature of this patch; it is mainly about shuffling some existing logic around.
@neil: You could maybe sprinkle some comments in there explaining why you are doing the lazy setup?
Nothing stood out from the diff in a quick review. I'll be test driving this branch until tomorrow and then do another review.
I think the code looks sane as such, the only problem I fear is some racy oddball edge case that relies on the old startup sequence. That said; I am not overly concerned by the nature of this patch; it is mainly about shuffling some existing logic around.
@neil: You could maybe sprinkle some comments in there explaining why you are doing the lazy setup?