How should Docker logging and observability be implemented for production?

Write structured logs to stdout/stderr, collect them centrally, rotate runtime logs, propagate correlation IDs, and monitor resource and restart metrics.