SQS polling_interval broker transport option ignored.

[acanpacan]

The polling_interval option documented in the Celery docs
https://docs.celeryq.dev/en/main/getting-started/backends-and-brokers/sqs.html#polling-interval seems to be completely ignored and sqs.ReceiveMessage is being called as soon as the previous call finishes.

Tested on Environment

• Python Version: 3.13.9
• Celery Version: 5.6.0
• Kombu Version: 5.6.1

Expected Behavior

The polling_interval decides the number of seconds to sleep between unsuccessful polls.

Actual Behavior

The polling_interval is ignored and next poll is done immediately after the previous unsuccessful poll finishes.

[auvipy]

is this the same thing with celery 5.6.2?

[acanpacan]

Yes, the issue is still there. Tested with celery==5.6.2 and kombu==5.6.2.

I did a quick debug and as far as I understand, the issue is that the SQS Channel implements it's own event loop and it always calls the next iteration as soon as possible regardless of how many messages were fetched https://github.com/celery/kombu/blob/main/kombu/transport/SQS.py#L624
When looking at the logs at DEBUG level - the subsequent calls are only separated by wait_time_seconds because of the SQS long polling mechanism.