Because of the characteristics of this test harness it will probably need to be defined as an external, non jtreg, test. The idea is to check the following, not a comprehensive list:

 - Types on the ASTs must be the same for both approaches just after the corresponding attribution approach ends
 - TA must execute the same unique type checks as SA during it's check phase, a different order is acceptable though
 - Propagation, or not, of stuck expressions to the corresponding enclosing context and their subsequent "unstucking" must be equivalent on both approaches