Well I think under this model, it would definitely make sense to use a dynamic resolution.
In the case where they were using a development profile that locked the gpu at a rate lower than 2.23Ghz, if they're using dynamic resolution, that resolution might sometimes be better on the dynamic clocks than on the fixed debug clocks. Which would be a 'pleasant surprise'.
I think it would be interesting to know, though, what the performance guarantees are or aren't wrt the debug locked frequencies. If it's the case that the dynamic performance is guaranteed to be at least as good or better, then I can see how that would be very appealing for a lot of situations, to work against the profile that most closely fits your bounds, and then reap any extra performance (or resolution, in a dynamic res situation) when you take the stabilizers off, so to speak.