Issue:
Making 3 identical API calls within 6 seconds to the same location returns temperature differences of up to 9°C and pressure differences of 22 hPa.

Example (Nov 7, 2025 ~07:40 AM PST):

• Location: 50.038417, -116.892033 (Kaslo, BC)
• Model: GFS
• Request 1: 2.0°C, 1028.5 hPa
• Request 2: -6.6°C, 1014.4 hPa (6 seconds later)
• Request 3: 2.3°C, 1036.0 hPa (6 seconds later)

Temperature variance: 8.9°C between identical requests

This makes the API unreliable for applications requiring consistent forecast data.

Questions:

1. Is this a known issue with load-balanced cache servers?
2. Can I request a specific model initialization time for consistency?
3. Would ECMWF model be more stable?

I've implemented retry logic with variance detection, but would prefer the API to return consistent data natively.

Any guidance would be greatly appreciated!

API Details:

• Endpoint: POST https://api.windy.com/api/point-forecast/v2
• Parameters: ["wind", "windGust", "temp", "rh", "pressure", "precip"]
• Model: gfs
• Occurs consistently across different times/days