A block explorer whose index can be explained and rebuilt

An explorer is a queryable view over one or more nodes, not an independent source of consensus. Its correctness inherits node software, node configuration and the branch observed by that node. The interface therefore reports sync height, peer status and index lag rather than presenting every response as current.

We do not promise that an explorer can correct an invalid or compromised node. Independent node comparison can expose disagreement, but protocol correctness requires review at the node and consensus layers.

A transaction can move from confirmed to pending or disappear when a stronger branch replaces indexed blocks. The indexer records block ancestry and the mutations attributable to each block so address histories, balances and UTXO spend states can be reversed in a defined order.

Replay starts from a verified checkpoint and uses idempotent writes to avoid duplicated transactions or totals. Deep reorganisation policy is explicit: automatic rollback may stop at a configured boundary and require operator review.

Address balances, transaction counts and rich lists are derived conveniences. Scheduled controls compare indexed tips, block hashes, transaction samples, supply totals and UTXO aggregates with node data. A mismatch marks affected views as degraded while investigation or rebuild proceeds.

An index that has silently drifted is worse than no explorer because users may act on plausible but false data. We favour visible unavailability over serving results that cannot pass consistency checks.

Unbounded address histories and broad scans can exhaust database, cache and node capacity. Cursor pagination, maximum ranges, request quotas and separately budgeted exports keep ordinary lookup responsive. Cache keys include chain context so a rollback cannot leave canonical pages backed by orphaned data.

Mempool data changes more quickly and varies between nodes, so it receives short cache lifetimes and qualified labels. API consumers receive stable schemas and rate-limit metadata, but availability still depends on the explorer’s node and operating capacity.