yeah, it's a useless metric because it is noisy

what would be useful would be an average ping over a minimum 24 hour period

i could write a simple thing that would log the ping delays and make a EWMA graph that shows you its performance over time

the thing is that it's not just one thing... how much bandwidth this thing has, how many threads, which relay it's running, how many clients it gets, at what time, really a 24 hour sample is minimal to really evaluate... and measuring that against the bandwidth delivery as well, like, when you make big requests randomly across that 24 hour period, to count the time from request until the complete delivery... and probably you also want to in parallel have another evaluation of the quality of your connection to act as a baseline zero