Is the refresh target limited to the just the components of interest? If not might want to wrap them in a section and further isolate them to target the refresh section more appropriately. You can trace the processing of this sequence to confirm the main delay sources by reviewing the elapsed times.(likely) the section refresh. You could also use fiddler to see what specific requests take what amount of time to fullfill.
Use Pega tracer and fiddler trace (but probably not at the same time!) to clarify for yourself exactly what is happening when you change one field and tab out of it in the situation where you are unsatisfied with the performance. Does the trace show that more is occurring than you expect? Or does it show the right amount is occurring but it is taking too long? Depending on which you observe, your next steps would differ. /Eric