refresh_interval defaults to 60s

[nvtkaszpir]

Is your feature request related to a problem? Please describe.
How about both?
Sometimes we get very short living pods/containers, and seems like fluentd is not picking logs from them.

In fluend logs I can sometime see:

2021-05-13 13:25:41.729347863 +0000 fluent.warn: {"message":"[fluentd-containers.log] /var/log/containers/REDACTED-integration-tests-2_wait-for-postgres-6b84d958cbd73c933bc63757c224c9e73d4d385e190ba3345c4920e09d9bb9a7.log unreadable. It is excluded and would be examined next time."}
2021-05-13 13:54:41 +0000 [warn]: [fluentd-containers.log] /var/log/containers/REDACTED_wait-for-postgres-3be0321565f050b34a9cdf078e2e67fe8b7c827f0d946534bb14b9ddee53b883.log unreadable. It is excluded and would be examined next time.
2021

and I know those init containers can live very shortly (like seconds).

Describe the solution you'd like
in_tail plugin has option refresh_interval which is default 60s

Describe alternatives you've considered
refresh_interval 1s (kind'a mental, would keep it around 5s)

Additional context
In theory inotify should be sufficient, but looks like not?

Anyone experienced this, or experimented with setting refresh_interval to value lower than 60s?