When running an active node (BFT leader or stake pool) it is interesting to be able to make choices on how to manage the pending transactions: how long to keep them, how to prioritize them etc.
mempool field in your node config file is not mandatory, by default it is set
mempool: fragment_ttl: 30m log_ttl: 1h garbage_collection_interval: 15m
fragment_ttldescribes for how long the node shall keep a fragment (a transaction) pending in the pool before being discarded;
log_ttldescribes for how long the node will keep logs of pending/accepted/rejected fragments in the pool; This is link to the data you receives from the REST fragment logs end point;
garbage_collection_intervaldescribes the interval between 2 garbage collection runs: i.e. when the node removes item (fragments or logs) that have timed out.