This relates to a JISC funded project (Managing Research Data stream) called JoRD which is looking into this area - in the Centre for Research Communications at Nottingham University.
Melanie, the biggest challenge i face is being time-poor. if such a service would enable collaborative research with bigger sample sizes, then i am definately an affirmative. H'er, i have in the past shared datasets, and never heard anything again from the folks i thought i was collaborating with. we have had a few RCTs, where sample size was such an issue, we did not recruit enough clients to either the intervention or the control group, and so had no statistically significant results, but really useful clinically significant messages. the one in poarticular i am thinking of was an RCT with young children with ALL. great clinical learning, but nothing statistical.