I've configured and also MODIFIED Mondrian sources to achieve the extreme privacy level. I read Mondrian reuses cached data to build results, so if two users U1 and U2 share the same connection, data privacy is potentially compromised.

Then, in my environment, each user has his own database JDBC connection and his own Mondrian schemes even if they are exploring the same XML schema file than other users. In other words, the same schema is loaded as many times as users are requesting it.

Finally, I've modified Mondrian connections to ensure each of them uses the right underlying JDBC connection belonging to the user exploring the schema.

well, from the architecture point of view, everything seems to work as expected but I do not have enough data and resources to perform extreme performance tests.

Might anyone please anticipate those problems?