thanks for the question, they definitely make sense. I wished we had the numbers yet, but so far we are still under heavy development working on Alpha releases. Those numbers are likely frequently changing.
As a beginning of an answer, the collection of metrics will really depend on the infrastructure you are ready to put in place, by experience with RHQ, some will prefer to decrease collection or metrics to monitor to limit the management solution to a single host, some will want to have more precise datapoints (or be alerted sooner) and will invest in a larger deployment, IMO in general the collection would be between 5s to 5min for metrics that change often and 30m or 1h for metrics that barely change.
Question about disk usage and Cassandra cluster size are questions I really want to be able to answer ASAP, your question will help prioritize this. We need to tell the maximum ingestion rate for a single "average" server and how much disk space does it use over a year (extrapolating), unfortunately I don't have those numbers yet and we need to make this automated since it will change rapidly in the coming months.