[Sad Catapella N8][Scientific reporting] range_check_issues silently combines incompatible range settings and coerces channel identifiers #59

Baseline and classification

Final normalization audit Batch 8 at f784c95e3049483f87466150ffc5764375fa631d. Confirmed source-level reporting/validation gap, not a demonstrated corruption of source histograms. No package tests executed.

Evidence

• range_check_issues() lines 204–254 groups inputs only by int(channel), which silently maps True, 1.9 or '1' to C1. It does not check the channel identifier is nonnegative integral (cf. channel-ID integrity issue [Sad Catapella N3][Data integrity] Parameter channel indices silently truncate fractions and booleans #38 in fitting).
• In each channel the routine chooses first = found[0][1] and reports its central_percentiles, robust_quantiles, histogram_range and min_central_bins for every included plane, although each RangeCheck has its own values (and the public API accepts a heterogeneous iterable).
• RangeCheck lines 142–184 contains no __post_init__ validation. Existing test supplies one consistent setting and integer channels only.

Deterministic example

Provide two narrow checks in channel 0: plane A with histogram_range='minmax', robust_quantiles=(0.01,99.99), min_central_bins=100; plane B with histogram_range='robust', robust_quantiles=(1,99), min_central_bins=200. The function emits one warning describing both as if their thresholds and causal explanations matched the first plane. Reversing input order changes the scientific interpretation of the same evidence. If a caller passes channel 0.9, it becomes C0 silently.

Impact

Misleading QC advice about when robust binning would help, inconsistent threshold provenance and wrong channel labels for malformed input. This is not a claim that preprocessing currently mixes settings in one normal run; caller-side homogeneous settings may prevent the issue there.

Fix / acceptance

Validate channel as a nonnegative integer (reject boolean/fraction/string). Validate RangeCheck field consistency and group by homogeneous settings signature, or reject mixed settings with plane/channel references; retain one warning per channel when settings truly match as fixed by N60. Do not silently pick a 'first' configuration as representative of incompatible measurements. Test mixed histogram_range, robust_quantiles, central_percentiles and min_central_bins, permutations of input order, valid group, malformed channel, and stable issue text. Numba has no role.