Just curious if there is any guidelines in terms of expectation of response times if a blast of requests is sent to the TRKD servers? Also, if I have a requirement where I know I may need to issue a batch of requests to TRKD where it may impact normal performance, is there a configuration option per client based on expected load?