Hello, everyone. I've been trying to resolve it myself for a couple of month, but finally reached a deadlock. So I rise my question here.
For some HTTP(S) software services DCRUM intermittently detects abnormally high "idle time". Experimental validation and network trace manual studying of several cases shows that DCRUM considers real-user reaction time as "idle time" improperly linking two independent HTTP requests.
During my analysis, I found that DCRUM considers "Referrer:" field of HTTP request header as a factor to unite two neighborhood requests into a single "Operation" and measure time lag between them as "idle time". Also, I found that extending "content-types" configuration with all the content-types available at the affected HTTP application might significantly decrease the number mistakes. I could even resolve the issue for some of the affected software services. But still there are several ones, where this doesn't help to completely resolve the issue (see attached screenshot).
Questions: - Is there anyway to completely disable idle time detection at DCRUM or disable it for particular HTTP-based software services?
- What is a complete list of factors DCRUM considers to unite two independent HTTP requests into a single "operation"?