Cleanup from a review of the last three weeks' changes: dead code removed, duplicated operations folded into their owners (periodic-axis detection, owned-interior bounds, the run's CSV records, the particle cell mean, named-source resolution), and stale comments corrected. Results are bit-identical on every case compared, by checkpoint and by particle ID, at several rank layouts and both index widths. Defects fixed along the way:
- an analytical-field restart reports errors from the field evaluation instead of dropping them, and a non-finite contravariant or curvilinear cell source value stops the run;
- an inlet's synthetic-fluctuation stresses are refused unless Cartesian and unconstrained, as the CLI already ensured;
- with enforce_seam_flux, the driven periodic trim on the -Xi and -Zeta faces uses the boundary face's own area;
- Profiling_Timestep_Summary.csv keeps its header when its first row comes after step 1, and the particle snapshot table prints 64-bit particle IDs correctly;
- the CLI checks named sources and multi-block inlets as it does inline and single-block ones, refuses an infinite scale, validates the reducer's half_height, reads fold: "false" as false, refuses source constants named like expression variables, and summarize --table no longer parses a text log as CSV.