No database. No indexing pipeline. Every value below is fetched live from the ChronoTrie node at the moment you ask — exploiting its blocknum-keyed state to answer "state at block N" queries in-band. Contrast: a normal explorer (BlockScout) first copies the entire chain into its own Postgres and serves from that copy.
Enter an address. We call eth_getBalance(addr, block) at many
historical blocks — each answered directly from versioned state, no diff replay.
Backend: a ChronoTrie geth node on the blockpath (blocknum-keyed) archive scheme.
This page holds no state — reload it and the "rows stored" stays 0.
The counter above only ever counts live JSON-RPC calls to the node.