Most answers, and maybe the question, don't address how "scooping" normally happens. It isn't that group 2 simply takes group 1's results and publishes them. It's that group 2 realizes from group 1's partial results that there's some big finding to be had--or already made and just needing enough confirmation--and can take the already-gleaned insight to drive a frenzy of their own experiments (which in some cases are exactly the same as experiments being done by group 1).
Group 2, after having swooped in at the end to finish up really fast, then publishes without reference to how they realized these were the right experiments to do, and without any citation to group 1 since they're still finishing up.
Now, if all research were totally open, this maneuver would no longer work, since you'd be able to track that key ideas 1-5 were out there before group 2 jumped in with idea 6 that crossed over the threshold needed for a good publication. Even if group 2 got the good publication, it'd be really obvious that 80% of the valuable work was actually done by group 1.
But if only some research is totally open, then the closed guys can just feign ignorance about "scooping" the other group. They can probably rightly claim to have been working on the same problem for a long time, and that they weren't really doing the right experiments or thinking about the problem the right way is completely obscured.
Now, although I haven't witnessed this kind of scooping with open data, I have witnessed it (at various distances) on at least four occasions with data that was not fully open but openly shared in various contexts (meetings, etc.). And though most people intimately familiar with the details agreed that it happened, group 2 was the one with the better publication record. I think all the group 1s ended up okay w.r.t. funding and tenure and so on, but it was to their detriment.
So the concern seems potentially valid, at least until hiring and tenure and grant committees will look more favorably on a Figshare etc. or GitHub repository containing documentation of discoveries than on a high-profile publication months later that contain mostly those same discoveries.
This post has been migrated from the Open Science private beta at StackExchange (A51.SE)