The only best practice we have around exception handling is to make sure that if you are doing a "retry" on the failed operator that you have a reasonable exit path for the process after the retry. If the process continues a retry without ever exiting (or failing) the process, your environment will have a lot of looping processes that may never end which can impact performance greatly.
Other than that most users setup a 3 or 4 retry scenario for web service failures with 20 - 30 minutes delay between the tries. It really depends on what normally would cause a failure in your environment. If you know that the application that PAM is making the web service call to would normally be down for maintenance for hours at a time, you might want to set your exception handling to not try again for an hour or more.
There is an example exception handler process in the Predefined PAM processes that come with the Out-of-the-Box processes. It is in the Getting Started folder. Keep in mind it is a very simple example. Just to show you how it works.