Hi Alex,
OPES_METAD builds its bias starting from the probability estimate along the CV obtained from reweighting, say P(s_1). Using EXTRA_BIAS, one correctly reweights the bias due to another plumed action as well. If this extra bias is a function of a different CV, say s_2, then the reconstructed probability is no longer P(s_1), but something not well defined that is a function of s_2 as well. There might be corner cases in which this makes sense, but I can't think of one right now. If you really need this extra bias on s_2 and also need OPES_METAD to be aware of it, probably you should also bias s_2 with OPES_METAD. It makes perfect sense to use an extra bias that is a function of a subset of the CVs used in OPES_METAD.
Sometimes I found myself adding a wall potential on a secondary CV in order to avoid visiting some non-interesting metastable state (e.g. defected crystals). In this case however, you don't want to reweight for this extra bias, you want a clean FES without noise from these excluded states.
Hope this helps.
Best,
Michele