[Sad Catapella B4][Scientific QC] Technical-repeat counts include excluded-only folders and omit explicit single folders #82

Baseline and classification

Sad Catapella Batch 4, commit d7a538e56d41ac1093ba79adcad397c21b3dfd32. Confirmed counting inconsistency by source inspection; end-to-end fixture not executed.

Evidence

• biological_repeats() lines 173–186 counts included images with sum(image not in excluded ...) but increments technical_repeat_count whenever leaf.technical_repeat != SINGLE, even when all images of that leaf are excluded.
• The same expression counts a real directory named single as 0 even though discovery explicitly supports that directory at technical-repeat level.
• Contract §11 says only analysed images count and the balance table holds the number of technical-repeat folders. BalanceCell and DesignRow propagate that number into reporting and saved design.
• Existing test covers the case where the entire biological repeat is excluded, but not where one of multiple technical folders is fully excluded.

Reproduction case

ctrl/rep1/techA/a.tif, ctrl/rep1/techB/b.tif. Exclude techB/b.tif only: the included image count becomes 1, yet the displayed technical-repeat count remains 2 instead of 1. For ctrl/rep1/single/a.tif, the real technical-repeat folder can be shown as 0. Strictly separate the count of existing physical directories, counted directories with included images, and missing levels, then document which is displayed.

Fix/acceptance

Specify that a technical-repeat count is of folders contributing at least one included image if it is supposed to describe analysed data; derive it from active leaves rather than counting every leaf. Decide and document whether a literal single directory counts as a physical technical-repeat or represents a logical singleton level with count 0; keep raw path provenance so reporting cannot conflate them. Add exclusion and explicit-single tests in dataset, run and confirmed-design round-trip.

No Numba value: correctness of grouping rules first.