The problem with taking a snapshot is that it quickly becomes out of date, and as you mention can be different with different content. We use some fuzzy selecti

The problem with taking a snapshot is that it quickly becomes out of date, and as you mention can be different with different content. We use some fuzzy selection methods to select the underlying DOM node that is causing you a problem. When a user goes to look at that bug, they're taken to the actual page and location where the problem occurred. This means that even as the page changes our bug tagging still points to the right place. We also keep tabs on all sorts of user/browser information to ensure a bug can be replicated. We have some more tricks up our sleeve for replicating page state and user data (not yet released).