In PDB-5021 we found an edge case in the way a factset's package_inventory was included in the hash for the factset. The order of the package_inventory mattered in the hash calculation and it was possible with sync that two otherwise identical factsets with differently ordered package_inventory vectors would generate different hashes. This caused sync to repeatedly pull a these factsets until another was submitted.
We should audit the hashing for the other entity types to double check that there aren't similar situations where the ordering of a piece of an entity influences the hash that's calculated.