I have found that anchored results tend to change less over time. Whether this is good or bad can be debated.
Since anchored results also include all past data, if you have market disrupting event in the data (think 2008-9 Financial Crisis), the effect of that will always be present in future optimizations. With non-anchored, that event eventually disappears. That can make a huge difference, and again, could be good or bad.
One other point is that you need to know how the walkforward engine processes results, since it may incorrectly calculate some fitness functions, depending on how it operates.