In none deterministic sequential derivation mode only one rule application is performed in each step. If several applications are possible, one is chosen pseudo-randomly.

Assuming a set of two rules A ==> B A ==> C is applied on an axiom A A in sequential_mode the result after the first step could be B A or C A and after the second step B B,B C,C B or C C. Because at any step only one match is processed an the rule is picked at random.