I'm literally scanning the `mempool` table on a single Stacks node. Consider that different nodes on the
network may receive different transactions in different orders. So the mempool on one node may be different
from the mempool on another.
Related to above: the Stacks node is very conservative about removing transactions from this table. So
transactions already processed might still linger in it, and transactions that might never be processed
(e.g. fees are too low or there's a nonce gap) might remain in there until they are aged out. So the actual
number of transactions that can be processed is likely much lower than what's reported here.