I am not aware of any published best practices for this sort of monitoring. Generally speaking, thresholds are an individual/personal choice depending on the needs of the user and/or the business. For example - some users might consider 50% CPU consumption to be "high" on a server and require an alarm for that, while other users may not be concerned until the CPU reaches 80%.
Generally speaking, a good practice if you are not sure what thresholds to use is to establish a sort of baseline - let the probe gather QoS for a week or so and take a look at the average response times, and then you can base your thresholds on this (for example, if the normal response time appears to be 10ms, and varies up to 15ms sometimes, you might consider alarming at 20ms.)
Perhaps some experienced AD Administrators can comment about what thresholds they are using but we do not really have any detail like this in Support.