Hi,
We have a L4 quota of 3 in europe-west4, the quotas page shows us exceeding the quota, peak utilization at 5, while we observe no api errors or QUOTA_EXCEEDED responses.
We have looked through the available documentation but there are no satisfactory answers. From our testing, we can get 4 parallel executions going with a 5th one marked as pending until capacity is freed up. However this contradicts the documentation around quotas where there are explicit warnings about dropped requests if quotas are exceeded.
Thus I ask, do cloud run jobs manage their quotas internally, and throttle to not exceed them, or is there a risk of failed requests if we queue a significant amount of jobs over the quota (10 for example)?
Our workloads are both time-sensitive, come in bursts, and are sporadic, thus we have been looking at cloud run jobs as opposed to worker pools or services.