The main idea: We scan over all the documents, hashing them, multiplying hashes (order-independent check, xor can be used as an alternative) and then using the same logic as in basic verification to combine the resulting hashes with the respecting namespaces and get the final hash.

The cons of this approach -- it takes the same amount of time as the reading of the whole DB. There are different ways that we could use to improve the performance (e.g. hashing on the server side using js functions), but this approach is not universal and should be tested across different DBMS separately. Probably, an optimization, that can be implemented in the future.

The code is a draft, which means:

• There are no tests;
• A lot of boilerplate (was planning to clean it later);
• The command line key is worded horribly (--verify-fully).