When fluentd server is down and async=true, the logger will try to reconnect indefinitely

This causes Docker container to hang: https://github.com/moby/moby/issues/40063

Start by reproducing the async=true behavior with the Fluentd server unavailable, then trace the logger's reconnect path; no source files or tests are named in the issue. Compare the behavior with the linked Docker report and establish bounded reconnect behavior that allows the container to shut down, with tests covering the unavailable-server case.

Written by the indexing model from the issue text.