Sometime after that, looks like the next day, both downloads were pulled, allegedly due to an index corruption issue (if you believe the commentary on the Release Services blog post for CU #6) or due to an issue with w3wp.exe (if you believe the label on the hotfix posted directly to the CU #6 article itself). Or perhaps both. At the very least you may see a regression involving index corruption when you rebuild a clustered index online and the plan involves a non-clustered index. I don't have a repro handy.
On June 13th, both articles were updated with new versions of the cumulative updates that no longer exhibit this symptom (sound familiar?). So if you download today, you'll actually get:
Please check @@VERSION. If it reports 10.00.5840, 10.50.2874 or 10.50.4279, please go re-download and re-apply the relevant cumulative update for your branch. If you're on 2008 R2, please see my post where I announce new CUs for the 2008 R2 branches.
In cases like this, I really, really, really wish they would release a new, out-of-band cumulative update (this is a non-issue, now, for R2, but remains an issue for 2008 SP3 customers for another month or so). How many people go back and take another look at the CU article once they have downloaded, installed, and regression tested? Does everyone regression test a full set of online rebuilds that probably don't occur that often? I think many users will simply not see this update and are on the verge of experiencing the symptom – there simply isn't enough visibility when they quietly swap out the versions that the download links yield.