Unless my calculations are wrong, executing a NOP (opcode $00) would take 39 cycles (instead of 4 cycles on a 8080).

Here is a simulation of an execution of NOP from the label run_emulator in 8080.s with STACK_BASED_DISPATCHER=false (it's 6 cycles faster)

  (cycles, instruction)
  3 lda (PCL),y
  2 tax
  2 asl
  3 bcc do_tab1
  4 sta _jmp2+1
  4 lda instruction_length,x
  3 beq length1
  5 inc PCL
  3 bne no_inch_pch
  5 jmp (tab2)
  3 jmp run_emulator
  =39

And that's in the best case with no page crossing.