Yes, it seems that the FF initial state are in the metastable state which requires some form of initial input (initialization/boot) to stabilize the FF out of the metastable state and into a known state (either on or off). I notice that trying to boot from the first FF and allowing it to fill all 8 FF will take forever so, I devised a parallel boot sequence setting two separate groups of FF, hence the piece of wood in the mid section of the layout which blocks the laser after the boot sequence. I could have extended the parallel boot system to 4 groups of FF but my laptop performance wasn't up to the task.

I have been working with laser logic as well in an attempt to construct a BCD counter using 7-segment displays. I have been able to build D FFs and SR FFs, but I have been running into metastability issues with T FFs and JK FFs. Decade counters require little additional logic with a T FF implementation. Is the initial boot sequence used in order to overcome the oscillatory nature of the FFs in your circuit?