In most cases the search and replace regex will not match between different inputs. Usually the regex used is also not complex enough to see any slow down by having it run without finding any matches.
pdttohtml only runs once per PDF input. Each search and replace expression is run over the html produced by pdftohtml. It actually runs over the html produced by any input plugin. Each expression is run once per page. In the case of PDF input only one page is generated so it is only run once. In the case of an EPUB book (or an input plugin that creates multiple pages within the OEB) there is no way to avoid running the search and replace multiple time. It has to run over each independent page.
If you're seeing the search and replace run multiple times it's because the document has multiple pages and it has to be run over each page. When I say page I'm talking about the independent xhtml files found with an OEB (EPUB).
Also, there is a bug I'm aware of with the regex input field. It should be saving all previous regex used in the drop down. It's not and I plan to look into why. Once that's fixed you can put your expressions, save, then remove it, then later find it in the list to reenable it.