Replay & resume

Slipstream keeps the last 300 slots, about two minutes, for reconnects and for a consumer that falls behind.

Resuming after a reconnect#

Record the slot of the last update you processed. When the stream drops, reconnect and send the same request with fromSlot set to that slot: the stream replays everything your filters match from that slot onwards, then continues live. Process updates idempotently, since the first ones may repeat what you saw before the drop.

Outside the window#

A slot older than the window cannot be replayed, and a request for one is refused. After a longer outage, rebuild your state from another source (an RPC node or your own store) and subscribe again without fromSlot.