Persistent provider budget exhaustion (OpenAI insufficient_quota,
surfaced as an HTTP 429 whose body carries that code). Distinct from
the transient LLM_RATE_LIMITED throttle: quota exhaustion is
fail-fast (non-retryable) and is the signal a global AI-budget
breaker trips on.
Persistent provider budget exhaustion (OpenAI
insufficient_quota, surfaced as an HTTP 429 whose body carries that code). Distinct from the transientLLM_RATE_LIMITEDthrottle: quota exhaustion is fail-fast (non-retryable) and is the signal a global AI-budget breaker trips on.