CollectionCacheRecoverer::recoverCollectionFromDisk always rebuilds the full routing table from the authoritative shard catalog by passing ChunkVersion::UNTRACKED() to makeCollectionAndChunksAggregation, so every recovery reads and parses all of a collection's chunks.

When the shard already holds a routing table that was just invalidated, we should instead pass that table's collection placement version as sinceVersion, so the aggregation returns only the chunks changed since then and merge them onto the existing table via RoutingTableHistory::makeUpdated , mirroring how the catalog cache loader does incremental CSRS refreshes.