Noticeable 503 errors and latency difference between older 5months vs newer 2months paid API keys

Gemini API
,
1

Hi everyone,

I’m running into a frustrating performance gap between two of my Gemini API keys. Both are on the paid tier, both are calling the exact same model under identical workloads, but they behave completely differently:

  • Older Key (5 months old): Extremely stable, low latency, and very few 503 errors.

  • Newer Key (2 months old): Constant 503 Service Unavailable / High Demand errors and significantly higher latency.

It seems like my requests are being routed differently based on the age or maturity of the underlying Google Cloud project.

I wanted to ask the community and any Google team members here:

  1. Does account/project maturity (historical spend or account age) affect our backend priority queue or “Usage Tier” routing?

  2. Are requests from newer paid projects subjected to heavier traffic-shaping/throttling during peak server loads compared to established, older projects?

  3. Has anyone found a way to manually sync, migrate, or upgrade the tier/priority status of an older project over to a newer one so performance is consistent?

Appreciate any insights or similar experiences!