Is your feature request related to a problem? Please describe.
When you have the AWS credentials file which is under rolling update by some scheduler then after some time fluent bit still working with old AWS credentials and failing with 403 when sending metrics to AWS Prometheus.

Describe the solution you'd like
fluentbit should reload these credentials out of the box.

Describe alternatives you've considered
I've used batch job which is restarting the service itself or did a hot reload

Additional context
In case if users have rolling credentials then they should prepare workarounds.