For counters, I used one of the counter plugins and then feed that into another one that does percentiles on a longer time sequence. These are saved for all of our apps. They go to another tag along with the name of the app.
For monitoring, we use Nagios with Python scripts, so I wrote a couple of those. I also have a script on each server with Fluentd that reports in every few minutes, and this goes to another log. This is useful because we're on Amazon with dynamic instances, so the number of Fluentd instances changes across the day.