Right now, if you pick a very large number of hosts, the loop that generates hosts will just loop forever, since there’s a finite universe that’s available.
The views that you get when you subscribe to the data stream are kinda too big now, if you look, for example, at the hosts with failures. We should make the arrival rate of failures smaller, so we don’t just got a big constant fraction with failures.
Since we replay the sequence upon start up of the client, lots of checks will initially look stale (compared to [Time.now ()]) until the client has caught up with the stream. If we exposed a notion of [`Caught_up] in the sequence, we could silence this noise. It’s probably not worth it though.
yminsky: You can just not display things until time has roughly caught up with real-time, e.g., within a second.