It was so far based on the order in which algos with identical names were declared, leading to major issues with ref updates. It now very simply uses the already existing "identity" of the algorithm (a hash of its config basically), which is unique
@jonrob @rmatev @tnanut This should tackle #267 and hopefully solve the issues found in !2773 (merged)
Part of set !3872 Rec!3196 Allen!1072 Moore!1909 DaVinci!790 MooreOnline!190 MooreAnalysis!100 Alignment!333