Partner models (Gemini/Nano Banana 2, GPT Image 2) fail with "Something went wrong" after 3-4 generations while Firefly native models keep working - looks like a rate-limit issue, not out of credits

Account: yansharetto@gmail.com

Summary: All Partner model image generators (Gemini 3.1 Nano Banana 2, GPT Image 2 tested) fail with "Something went wrong. Please try again later." after roughly 3-4 successful generations in a session, both yesterday and again this morning. Native Firefly models (e.g. Firefly Image 5) keep working fine on the same account/browser throughout.

Console error each time: "Catch Colligo error: 408 timeout_error system under load"

Steps to reproduce:

1. Generate image, prompt "gums", no reference image, default settings.

2. Model: Gemini 3.1 (Nano Banana 2) - fails after a handful of generations.

3. Switch to GPT Image 2 - same failure, same console error.

4. Switch to Firefly Image 5 - succeeds immediately every time.

Already ruled out:

- Not a credits issue: 3935/4000 Premium credits remaining, resets Sept 22, 2026.

- Not browser cache: full reload plus clearing local/session storage did not change the outcome.

- Not model-specific: reproduced identically on 2 different partner providers (Google and OpenAI).

- Checked status.adobe.com directly - no open Firefly incident listed.

This looks like a per-session/short-window rate limit on partner-model calls that is returning the same generic error as a full outage, rather than an actual capacity outage - please check backend throttle logs for this account around the timestamps of the failures. Getting only 3-4 partner-model generations before lockout, on multiple consecutive days, is blocking real work.