During cluster installation, assisted-service generates a stream of events. Events are associated with progress on a particular host, not with progress of the cluster installation overall.

The installer's agent wait-for bootstrap-complete subcommand polls the event list, but only prints out the most recent event (as long as it is new). This means that all we see in the logs is the progress of a randomly-chosen node (a different one each time we poll). Important events like errors are not likely to be seen because there is a high probability that the latest event is from another node.

It would be better if this worked more like tail -f, where we start with the latest events but then print all subsequent events.

Or perhaps in the spirit of the original idea, we should log the latest event for each host (if it is a new event we haven't logged before).