What is the recommended pattern for calling the Snowpipe loadHistoryScan endpoint to avoid rate limits?

• A. Reading the last 10 minutes of history every 8 minutes.
• B. Reading the last 24 hours of history every minute.
• C. Reading the last seven days of history every hour.
• D. Reading the last hour of history every 15 minutes.

Answer: (A)

Polling strategy should balance data freshness with rate-limit safety. The best approach uses a small, sliding window and a cadence that keeps the window overlapping with the next poll. Reading the last 10 minutes of history every 8 minutes does this: each call grabs a short slice of the most recent history, and the 8-minute cadence ensures continuous coverage with overlap. If some history arrives late, it’s still captured on the next poll within a short grace period, while keeping the number of calls and the data per call moderate to avoid hitting rate limits.

Other patterns are less suitable: querying a 24-hour window every minute creates excessive traffic and data volume, quickly exhausting quotas. Querying the last seven days every hour is too slow to surface timely events and risks missing late-arriving history. Querying the last hour every 15 minutes is better than the extremes but still yields more latency and larger per-call payload than a small, overlapping window with a tighter cadence.