Batch jobs sent to GPT-5 keep getting rejected due to "1,500,000 enqueued tokens" limit even though no other job is running

Hi,

So, as in the title, I am trying to send some batch jobs to GPT 5. Actually 4 out of 6 batches finished successfully. However, the last two keep getting rejected (already for many days and attempts) with “1,500,000 enqueued tokens” limit error. What is weird about it is that I have no other jobs running, so the error seems weird and not exactly accurate.

Any ideas for what can be done about it? If that is important - I am on tier 1 access.