The width-search range controls the reference flux
The owner chose more numerical refinement before deciding how the MAST-U reference should treat its small negative heat fluxes. One bounded diagnostic extended the final basis to 16 functions and widened TGLF’s width search. It completed, and its saved outputs pass an independent audit with all controls reproduced (3029 integrity checks pass on the complete 78-pair extension and all 36 control comparisons reproduce job 2687308; NBASIS_MAX 16 still moves electron flux by up to 0.48 gB per step, and widening the width search from 1.65 to 3.3 changes it from about 1e-3 gB to as much as 181 gBC-206).
Neither change settles the flux. The near-zero values seen at the previous largest basis disappear again at larger bases, and the last step still moves the electron flux by up to 0.48 gB. Widening the width search matters far more: the default range turns out to be binding, and letting the search move past it turns a residual of about 1e-3 gB into as much as 181 gB at one state. At the standard settings, the two negative values become clearly positive.
Across three bounded studies the “refined” flux at these states spans more than five orders of magnitude. The standard settings, which labelled all the training data, therefore act as a convention that defines the model here rather than as an approximation to a converged answer. How to use that convention for the reference goes back to the owner. The current model, guard and acceptance criteria are unchanged.
Technical details → Verification › reference extension results
Decisions → OQ-33 · OQ-35 · OQ-36