The script has a very small code-base, very fast, very low bandwidth and memory usage.
For this it relies heavily (and almost completely) on memcached, no data is stored anywhere, and it lets memcached do most of the lifecycle management.
Atrack also aims to respect your privacy: other than what is needed for the most basic tracking (hashes/keys and ip/ports) atrack gathers no information whatsoever.
Beyond that no aggregate statistics are kept of anything (at least once the testing/debugging phase is over), and nothing is stored permanently anywhere, not even hashes and ip/ports, they are kept in memory until they are expired by memcached and never written to disk or any other kind of permanent storage.