It seems that at some point in the past the minio instance running within podman started terminating repeatedly, causing intermittent downtime (as seen in the historical graphs).
It had no real reason to be containerised and thus the binary was run as a low-privileged user directly, and the problem ceased to occur.