We need to be very careful when we talk about how much RAM (virtual memory or physical RAM) an application or other component is using, because some designs will simply try to find a use for as much spare RAM as possible in order to improve performance, caching is a prime example (and not just disk caching). Caching the results of work that has been carried out saves having to repeat that work again. If done intelligently, exploiting RAM as much as possible comes simply down to making the best use of resources, but there can be an element of voodoo in getting the balance right, since having to kick out dirty pages to free up some physical RAM for an essential use has a cost and squandering RAM because of ‘optimisation’ strategies when other processes really need it does not make sense. Any such behaviour needs to be adaptive, flexible and well-tested in umpteen scenarios and can easily backfire.
A web browser might in theory save the results of rendering a web page, either just saving the results of the layout calculation or even going as far as saving the entire page image drawn, and might save downloaded cached pages in virtual memory rather than just on the hard disk. Rather vague, but many years ago, I got the feeling that in some circumstances Opera was keen to cache web pages more than IE6 in terms of responsiveness when hitting the ‘back’ button.