Publishing the number of leaf and branch nodes helps us understand quite a few things:
Branch-compression = (num branches * nodeSize) / (number of leaves * (keysize+8))
Leaf compression = (#leaves + #blobs) * nodeSize / uncompressedSize
TotalBranch size = num-branches * nodeSize
ratio of branches/leaves = #branches/#leaves.
Total branch size for a query/query set gives a good idea of a typical node cache size that would avoid branch misses (with scaling depending on whether it is decompressed).