We use them when we replicate very complex dependency trees. Often with 1e6+ heterogeneous objects in the tree. If we replicate a tree, say, for regression tes

We use them when we replicate very complex dependency trees. Often with 1e6+ heterogeneous objects in the tree. If we replicate a tree, say, for regression testing, we can trivially confirm that it replicated identically. If not, we can use the partial trees to, very quickly, triage the discrepancies to the earliest divergence point. We also have a number of cases where we want to quickly confirm that 2 (or more) trees have identical dependencies in parts of the tree even though the overall trees diverge substantially. A Merkle tree approach works nicely here too. Indexing doesn't help in these cases as what we're doing is, effectively, fast content matching.