This doesn't seem to matter that much, because the models with the lowest deviance in each year collapse to a single stratum, recapture probabilities don't change that much over time, and there are no strong trends in CPUE or mean individual length throughout the season.
https://github.com/commfish/seak_sablefish/blob/de92f1e7cba8cdf22288da7016be7894c3dbe47e/r_code/mark_recapture.R#L702 @ben-williams
This doesn't seem to matter that much, because the models with the lowest deviance in each year collapse to a single stratum, recapture probabilities don't change that much over time, and there are no strong trends in CPUE or mean individual length throughout the season.