The most interesting part here is that the change is been driven largely by young republicans. They went from being two thirds (65%) in favor of gay marriage to just under half (49%) and also underwent similar changes in their support for anti-discrimination laws.
An alternative interpretation could be that this is another example showing polls are having increasing difficulty getting good representative samples of young voters. This would be consistent with the crosstab issues that seem to be occurring in Presidential election polls. If there is a sampling issue, you would expect it to show up both in those polls and also in issue-related polls like this.
So it is interesting that apparently the source of the decline they measure is specifically from younger respondents.
Of course, it is possible that some (or even all) of the decline could be real. But this is an alternative interpretation and possibility as well.