We're most likely to bump into loops when we're computing the initial fixes, not when we're combining fixes. By using a hybrid where we use interpreted code for checking combined fixes but non-interpreted for the initialFixes, we can avoid loops.
I just want to say: finally, it works! The issue was too much parallelism in PseudoGenetic, namely that we would be compiling the same module at the same time, and it would sometimes get confused.
We're most likely to bump into loops when we're computing the initial fixes, not when we're combining fixes. By using a hybrid where we use interpreted code for checking combined fixes but non-interpreted for the initialFixes, we can avoid loops.