Pre-registration: A Forward-Looking Blind Test of a Setup-Aware Critical Heat Flux Model Against Version 6 Recalibrated Without Setup Values

This programme's premise is wrong if a critical heat flux (CHF) correlation that ignores the test setup predicts across laboratories as well as a model that takes the setup and its environment into account. This record pre-registers a test of that sentence. A setup-aware model is fitted to tube CHF data from many laboratories and compared with a setup-blind baseline: version 6 of this programme's water CHF blend, recalibrated on the same fitting laboratories without any setup value, so the two differ only in the setup terms. "As well as" is fixed as a 10% margin in laboratory-weighted RMS log error. Almost every accessible tube CHF data set from a laboratory outside the fit has already been seen by the programme, so the test looks forward. Every admitted source, older or published after this freeze, is obtained, sealed and transcribed blind. Frozen code then sorts each point into three tiers: a duplicate of a seen point, a new point from a seen source, and a new point from an unseen source. Only the third tier carries the verdict. A source key fixes, for 130 sources, which laboratory each belongs to and whether the author or the analyst has seen that laboratory or that document; a shared university or company counts as one group. The deposit holds the protocol (plain text and PDF), the source key, the seen point files in every quality variant they support, the five frozen scripts with their self-tests, the bibliographic candidate lists and a power simulation. The simulation is stated as a limit: with a modest setup effect the test has moderate power at best, and little if laboratories differ by about 15% (the largest spread simulated) for reasons outside the model. No test data has been read. Archive SHA-256: bf917ce5eb8b48d0821ab9b01d440b865f4cef94ad7b2ad3cd0cfe065d6ab159