Re: Replay runs very slowly very big response times
"ignore think time"
Poor option, unless debugging only. You want to represent a human user, not a machine that can respond instantly. Think time is time between keyboard and chair. This will overdrive the system reducing the chaotic interval between requests to zero.
"Scan for Correlations"
The manual methods never fail. As noted in training, record twice, diff, examine for correlation candidates, find in the source and correlate. Anything else is a silver bullet hunting for a rare werewolf.
"...page cannot be displayed"
It is possible that to display a particular component requires a browser add-in for rendering. The classical example of this is PDF display in a browser, but certainly other candidates exist.
"page does not exist"
Several items could be coming into play here. You may have missed a correlation, which is highly likely given teh model noted above, or you may not be checking for valid return pages and the server returns an unexpected page that goes unhandled by a change in script flow in your script.