For bring-up you will almost certainly get away with it, and stitching capacitors will help, but be clear about what they do and do not fix. A stitching capacitor gives the return current a path between the two plane sections at that point, so it limits the loop area and controls the crosstalk and the impedance discontinuity locally. What it cannot do is remove the inductance of the capacitor and its two vias, which is somewhere around a nanohenry all in. At a 400 ps rise time you are looking at meaningful spectral content past 1 GHz, and a 1 nH series path is roughly 6 ohms of reactance at that frequency — enough that the return is still partly finding its way around the split rather than through your capacitor.
The practical consequence for your prototype: functionally it will run, and you will probably not see it in the eye diagram at 100 MHz. Where you will see it is the EMC chamber, because the current that does not go through the stitching capacitor goes around the split, and that loop is an efficient radiator at exactly the frequencies where the limits are tight. Place a stitching capacitor within about 2 mm of each crossing point, use the shortest via stubs your stack-up allows, and treat the prototype as a functional vehicle only. Do not book pre-compliance on this board expecting to pass.
