Some of our system tables (e.g. table_status) perform operations that have to go over the network. They compute one row at a time, so if you have a slow network or one of the servers is busy, reading such a table can take a considerable amount of time.

We should probably parallelize this and compute multiple rows at once.