Problems with 8/A Programmer's Panel

Several years ago, Vince Slyngstad and I "cloned" the rare DKC8-AA Programmer's Panel for PDP-8 with some improvements (0.6 vs. 0.3" LED displays, real "click" buttons instead of that membrane keypad, and fixed a couple of bugs in the original design). Also, since 8235's were scarce even then, we used 74F244 buffers instead to gate the input vs. output of the switch register onto the bus back to the M8316 Option 1 board. That may have caused a lurking problem, since occasionally the SR will change bits when an address is loaded. Turns out an oscillation starts on the ground lines at about 20 MHz with certain entries at the keypad to LSR and LA (I can't see an obvious pattern). This parasitic is large enough (up to 2V p-p with respect to the chassis/rack) that it's clocking the D-flops used as the physical switch register! It can be seen in the 7-segment displays as a "flicker" which will change when I touch the ground on the board... attempts to add more grounds actually made the problem worse! I'm thinking of redoing the board layout with better attention to ground planes/power buses. There's a .01 uf decoupling cap at every IC and 12 out of 40 pins on each of the two connecting ribbon cables are dedicated to ground. 7 on one, 5 on the other are for +5 volts. On the other hand, I am tired of tinkering and considering just buying a real DKC8-AA if there's one out there. Anyone? ;) -Charles