Method recordSuccess

pkg/runtime/fallback.go:172–179

recordSuccess updates the per-agent cooldown state after a successful model attempt. If a fallback rescued a non-retryable primary failure, pin to that fallback for the cooldown window. If the primary itself succeeded, clear any existing cooldown (handles both a clean primary success and recovery af

(a *agent.Agent, modelEntry modelWithFallback, primaryFailedWithNonRetryable bool)

// succeeded, clear any existing cooldown (handles both a clean primary
// success and recovery after a cooldown expires).
func (e *fallbackExecutor) recordSuccess(a *agent.Agent, modelEntry modelWithFallback, primaryFailedWithNonRetryable bool) {
	switch {
	case modelEntry.isFallback && primaryFailedWithNonRetryable:
		e.cooldowns.Set(a.Name(), modelEntry.index, getEffectiveCooldown(a))
	case !modelEntry.isFallback:
		e.cooldowns.Clear(a.Name())
	}
}

// classifyAttemptError handles an error from a stream attempt: checks for
// context cancellation, classifies the error, and returns either a