The whole of observations and deviances get passed to oem inside goFish, at mp.R:220-221
ctrl.oem$observations <- observations(oem)
ctrl.oem$stk <- stk.om
It would make writing oem functions simpler if those two where subset before calling oem@method, so no datayears subsetting inside the oem function: stk, observations and deviances will finish in the last data year, as determined by args.