The two-flop synchronizer
You will learn
Why two flops are the fix, and what the generated model of the crossing looks like.
The fix is two flops in the destination clock, nothing between them: the opening one may go metastable, and the second samples its settled output a full period later. Each extra stage multiplies MTBF by 2 to the power T_clk times log2 e over tau -- 295400 in Q10 at 100 MHz, the spec's stage_gain_q10(). The recording shows t27c gen on clock_domain.t27: the printed Zig is the crossing model the whole course leans on, and wc -l counts it. What the two flops cost is latency: two destination clocks from edge to settled output, the same price the release pipe of module 4 pays.