There seems to be no explicit treatment in ASME Y14.5 of this situation. The assumptions are that there will be a single radius and that the radius surface is exactly tangent to the neighboring surfaces.
For a feature carefully made with a single linear cut by a good quality end mill that isn't too bad an assumption.
But if it is by using a profiling pass with a cutter that is smaller than the width of the nominal slot or if the radius ends are drilled in advance of the rest of the slot or if the feature is waterjet or EDM cut, there isn't a similar guarantee of the shape.
Even in examples of position tolerance combined with rounded end slots, the standard has no clarification.
One is stuck unless a profile tolerance is used, but that isn't friendly to the assumed variations in plus-minus control of slot length.