Hi Paul,
If I have understood what you are proposing, then it should not be done for submissions to the database as a single same image would contribute to multiple measurements/observations.
My understanding is you would first combine images 1-9 to get a good signal, then images 2-10, and so on up to 9-18. This is of course simplified as you would have many images. Image 9 would be used 9 times, and so the same photons would contribute to 9 observations and that would be wrong.
You can do this kind of thing in an analysis, as long as what has been done is stated. Basically it is a smoothing function. It just should not be used for submitted observations to the database. This leaves the choice of whether or not to smooth with the researcher.
Best wishes,
Andy