You should keep WFO enabled. The goal of the MRC is to validate the out-of-sample (OOS) equity curve produced by the WFO process.
Recommended workflow:
Perform a normal [Train] of your strategy first to generate the .par files.
Run MRC.c in [Test] mode. This validates if your out-of-sample results are statistically significant or just due to luck.
Running MRC.c in [Train] mode is even more rigorous as it also checks for "selection bias" (overfitting during the optimization process itself), but it is computationally very expensive and can take a long time with WFO.
