Potential issue/inefficiency with crossover causing various problems

-Issue with genetics causes unexpected strategy to be produced outside of the parameter bounds: 
https://roadmap.strategyquant.com/tasks/sq4_9768

-crossover is allowed to bring in variables from one parent and insert them ANYWHERE in to the child. So for instance, the level for the RSI trigger could become the deviation for the Bollinger Band of the child which is nonsense and is probably the cause of the issue above as well as causing many crossovers to fail wasting CPU and reducing efficiency of the migration/island system.



To solve these issues I suggest to allow only entire indicators including all it's variables to be brought in from one parent. Single variables could be allowed to crossover only if it is for the exact same use, for instance a SL modifier of one parent can come in to the SL modifier of a child and or a TP modifier of another parent can come in to the TP modifier of the child. Never should the RSI level become the deviation for a Bollinger Band though.