This will get sawed up into 16 bits, and various people can play with theirs.
Fast parts are often badly characterized, and layout matters, so you have to just try things.
Two long vertical cuts will slice this into three slabs, and each of them can be cut up into individual boards. This requires intensive topological planning.
One controversial idea: we have a d-flop that has clock-to-output delay of 200 ps typ, and reset to output of 300 ps. So if you connect clock and reset together and apply an edge, that should make a 100 ps output pulse. That might replace my shorted transmission line that got clumsy to lay out.