Reloading duplicate samples efficiently, wait_async confusion

In my instrument, you can load up to 16 samples. The instrument itself can be in one of three sample modes ("time machine", sampler mode w/ reverse on, or "wavetable") set from a ui_menu. Because there's no way to change the sample mode from the script, I have to have 16 * 3 = 48 groups, each with it's own user zone and loop parameters. The problem is that the samples could be several minutes long, so reloading the instrument with the '!' button -- and to a lesser degree, loading a single sample -- could take a really long time. Even when the sample paths are all set to "", it still takes a solid second or two. This is basically what the relavent code looks like:

on init
    ...
    define NUM_ZONES := 48
    set_num_user_zones(NUM_ZONES)
    for c := 0 to NUM_ZONES - 1
        set_zone_par(NI_USER_ZONE_IDS[c], ZONE_PAR_GROUP, c)
        set_zone_par(NI_USER_ZONE_IDS[c], ZONE_PAR_LOW_KEY, 0)
        set_zone_par(NI_USER_ZONE_IDS[c], ZONE_PAR_HIGH_KEY, 127)
        set_zone_par(NI_USER_ZONE_IDS[c], ZONE_PAR_LOW_VELO, 0)
        set_zone_par(NI_USER_ZONE_IDS[c], ZONE_PAR_HIGH_VELO, 127)
    end for
end on

on persistence_changed
    ...
    for c := 0 to NUM_ZONES - 1 // 48 iterations
        wait_async(set_sample(NI_USER_ZONE_IDS[c], !sample_paths[c / 3]))
        wait_async(set_loop_par(NI_USER_ZONE_IDS[c], 0, LOOP_PAR_MODE, loop_button))
        wait_async(set_loop_par(NI_USER_ZONE_IDS[c], 0, LOOP_PAR_START, 0))
        wait_async(set_loop_par(NI_USER_ZONE_IDS[c], 0, LOOP_PAR_LENGTH, 2147483647))
    end for
end on

Questions:

1. Is there any way to load each sample once and put it into 3 groups instead of loading the same sample 3 times? Because the latter triples my loading time. Would it make more sense to only load each sample into the current sample mode's group, and then load all 16 again each time the user changes the ui_menu?

2. What exactly does wait_async() do? I've read the manual and forum posts about it but I still don't understand what it does vs. just doing the operations without it. I ask because if I do those 4 operations in the PCCB loop without it, it loads significantly faster. And I realized that all this time I haven't been using it in the ICB for the zone parameters and it's been fine. If the idea is that it waits until the current operation is completely done before moving on to the next, isn't that how programming languages already work by default, running instructions serially?

Thanks.