The original motivation for the session system was that the program (afni) would assume that the xyz coordinate of all datasets in one directory would properly overlap physically, but that xyz coordinates of datasets from different directories might not. With that in mind, I called the directories "sessions" with the idea that all the data from a single scanning session on a single subject would naturally "go together".
Of course, these assumptions can be erroneous (as can any assumption, excepting the Axiom of Infinity).
I could modify the interactive afni program (the only place where the session concept is used) to put all the input datasets into one big session inside the program. This would let you manage viewing any way you want. Of course, if you input a lot of files, then the 'Overlay' and 'Underlay' chooser lists would get pretty long.
That, of course, isn't what Anonymous asked for (and why the HELL do you think that I'd be pissed at a polite request?). But it's what would be simplest to offer.