April 7, 2013 at 2:28 pm #21306
If flexing the board changes something, there’s still a bad solder joint– that temporarily opens or closes when flexed –or the chips are not inserted all the way.
The selected row is controlled by the 74HC195 demultiplexer chips. Since the problem is in the top half of the rows, the particular chip to worry about is U2. The fact that you are losing regularly-spaced pins further suggests that that the problem is on the input side. That also means that the problem could also be at the U1, on the connections towards U2.
And, you still have that column to fix as well– the likely issue is at one of the pins of the LED driver chips. Check near the middle-bottom of U4, in the area by C6. And, make sure that U4 is fully seated.