There's a combination of a couple of things that define the latency. One thing is: How much data is being sent (Publisher) or received (Subscriber)? We currently support 2,000 events per second. If you aren't exceeding that, then you'll just need to consider PI Cloud Connect's retrieval rate and your own network performance. PI Cloud Connect sends/receives updates on 30 second intervals. In most scenarios, this means you won't see a latency of more than 1 minute on the Subscriber. Of course if you're dealing with any sort of network limitation that could incur a bit more of a delay.
I hope that helps! In my testing, I really haven't seen a latency of much more than 30-60 seconds on a healthy system.
Retrieving data ...