1. For those that do not know: two things you must consider for accurate long exposures are:
A. The accuracy of your lightmeter in extreme low light. (Most are accurate to 30s @ f/5.6 /w 100ISO film (i.e., EV0), to give an example for this situation at least.)
B. Reciprocity failure by the film. That is, the film failing to respond in the nice curve it normally maintains. That is, one would think exposing film for 10 minutes instead of five would make up for going from f/5.6 to f/8 (that relationship is reciprical, there is reciprocity), but it might only make up for going from f/5.6 to f/6.7 (This is NOT recipricol, there is reciprocity failure)!!! This is what happens at extremely long exposure times and it gets worse the longer the exposure is. (It also happens at extremely short exposure times, but we needn't worry about that underwater). In color negs/slides you also must consider differing rates of reciprocity failure between the color layers which can cause unnatural color shifts, though this perhaps is less important underwater. Soooo....
It is far too troublesome to correct for the color shifts underwater, which is why, as blacknet said, you might consider black and white for underwater. On the other hand, colors are kinda silly in aqualand anyhow. However, the overall exposure must be corrected to compensate for reciprocity failure. Soooooo...
It's better to have films that won't experience so much reciprocity failure so quickly. Nobody wants to have to think about yet another variable so let's pick some films that don't suffer so much.
Good choices (all exposure data from manufacturer, you may need to alter for yourself):
Kodak TMax 100 TMX, as blacknet suggets (B&W: No exposure correction needed for exposures up to 1/10 seconds, only +1/3 stop exposure compensation needed at 1 seconds, only +1/2 stop exposure compensation needed at 10 seconds, only +1 stop exposure compensation needed at 100 seconds)
Kodak E100G E100GX (Slide: No exposure compensation needed until up to 120 seconds, green shifting after 10 seconds)
Fuji Neopan Acros 100 (B&W: No exposure correction needed for exposures up to 120 seconds and only +1/2 stop exposure compensation needed at 1000 seconds)
Fuji Provia 100F (Slide: No exposure or color correction needed for exposures up to 128 seconds and only +1/3 stop exposure compensation needed at 256 seconds and there will be some orangeshift)
I wonder what Fuji Velvia 100F will have for reciprocity characteristics?
Here is a chart with reciprocity failure info for several types of slide film, including Velvia 50 (many of the other emulsions are older... notice the Provia is plain old Provia and not the F version):
http://www.tfkp.physik.uni-erlangen.de/yp/personal/ralf/photo/reciprocity.html
2. Shouldn't backscatter be reduced by a very long exposure without flash? The lack of flash should reduce it outright. The remaning particulates that cause backscatter should be moving slowly through the water and therefor it will be blurred, perhaps to the point of not being noticed depending on the exposure time, right?