If you're reading this, please take one minute out of your day and vote for the following Connect item:
If you're really interested, take three minutes: run the steps to reproduce the issue, and then check the box that says that you were able to reproduce the issue.
Imagine that ten hours ago you started a big transaction. You're sitting there waiting for it to finish, and it's running, and running, and running. At some point, you notice that the drive with the transaction log has filled up, so you create a second log file. And the transaction is sitting there running, and running, and running. Is it still doing work? Or did it catch the low disk space issue and start rolling back? Wouldn't it be great if you could actually answer that question? Are you surprised to learn that you can't answer that question?
SQL Server currently has a few DMVs that are supposed to tell us the status of a given request, transaction, and so on. These DMVs are unreliable in the vast majority of situations. That means that we are unable to answer important questions like, "is my transaction still doing any work, or did it die three hours ago?"
The SQL Server team needs to fix this. End of story. Please vote and help me convince the powers that be to do the right thing.