[woutervdb]

Could you not tackle this problem by not allowing this mingling of 'next' and 'current' variables? You would only allow a next variable to be a computation of current variables. Not sure to what extent this would limit the application of the paradigm, though.

[brap]

I believe this is exactly what OP meant, so likewise I don’t think ordering is a problem in his proposal.

[aappleby]

You can allow mingling as long as you don't produce an evaluation loop. The existence of evaluation loops can be detected by the compiler (Metron does this already).