Each module should have a unit test for it. Currently these tests are written using verilog test-bench scripts, though this is not necessarily set in stone.
Each module should be tested for propagation delay, using verigpu/timing.py. The result should be added to a comment at the top of the document. If the module is modified, the timing should be re-run.
If the timing script fails to run, this is likely because of unitialized regs that are being used in the combinatorial always block. If you open up build/netlist/6.v, and search for always, you can usually find hints as to which regs were not initialized for all execution paths.
Running verigpu/timing.py (see above) will also output the area, as a multiple of the area of a single nand unit. This line should be pasted into the top of each module too, and updated each time the module is updated.
Each module should be tested using GLS, as well as using the behavioral tests. GLS catches issues with timing, and with conditions in if and case, which might not be caught in the behavioral tests.