Calling Gemini 3.1 API keeps getting 503 error even though Gemini service is fine

My n8n HTTP node sends a simple image generation prompt to Gemini 3.1 but 50% of the time it fails with this error:

Gemini API error: 503 - “{\n “error”: {\n “code”: 503,\n “message”: “This model is currently experiencing high demand. Spikes in demand are usually temporary. Please try again later.”,\n “status”: “UNAVAILABLE”\n }\n}\n”

I checked availability history and the service seems to be working fine.

Occasionally I also get: Gemini API error: timeout of 600000ms exceeded

Out of 6 calls, only 3 max succeed. It also takes over 30 minutes to run the 6 rows.

Any suggestions for troubleshooting? What should I look into? TIA