fallocated blocks past EOF & past parent node range OK?

I'm sorry I missed the conversation earlier this morning, but ewhitney let me know that the consensus was that for fallocated blocks past EOF, it's fine to have those blocks live outside the parent node's range in the extent tree.

This seems quite strange to me, and unexpected.

What is the rationale for this? Why would these past-EOF extents look unique in the extent tree, and not be covered by the parent node?

Thanks,
-Eric