[Jammy2211]

Chapter 3 had lost the Fit Problems motivation and parts of its introductory teaching content during consolidation. This restores that narrative and presents six consecutive tutorials, ending with one adaptive reconstruction tutorial combining the former tutorials 8, 11 and 12.

The Bayesian evidence explanation now uses the approved Goldilocks analogy and explicitly selects the highest sampled log evidence. Lens modelling explains the large parameter-space volume of incorrect solutions, refers back to Chapter 2 Tutorial 5's positions likelihood penalty, and displays Maresca Figure 1. The preservation audit restores positivity and mapping-support diagnostics in Tutorial 1.

Scripts Changed

• scripts/chapter_3_pixelizations/tutorial_1_pixelizations.py — restore missing explanations and diagnostics, with pointers to later mask and mass experiments.
• scripts/chapter_3_pixelizations/tutorial_2_bayesian_regularization.py — evidence-based coefficient selection and mesh-resolution comparison.
• scripts/chapter_3_pixelizations/tutorial_3_borders.py — approved border and mask narrative and experiments.
• scripts/chapter_3_pixelizations/tutorial_4_lens_modeling.py — approved modelling narrative, embedded Maresca figure and likelihood-penalty references.
• scripts/chapter_3_pixelizations/tutorial_5_fit_problems.py — restore compact-source fit failures motivating adaptation.
• scripts/chapter_3_pixelizations/tutorial_6_adaptive_reconstruction.py — consolidate overlay/Hilbert meshes, split/adaptive regularization and an MGE-to-pixelization modelling chain; retire tutorials 8, 11 and 12.

Regenerated notebooks, contents, navigation and catalogues accompany the scripts. Scans retain scalar evidence and preferred fits to bound memory. The two new morphology experiments declare ENV: full_datasets in their developer-only __Env__ sections because the small-data cap distorts their evidence curves. The notebooks omit these declarations.

Validation

• All six tutorials passed full-size numerical runs, including the 75-point adaptive scan; 198 figures captured.
• All 41 runnable smoke scripts passed across the initial 38 successes and three targeted repaired reruns. The reruns resolved the two dataset-cap failures and an optional environment dependency.
• All 66 approved prose blocks preserved; early tutorials audited against repository history.
• Notebook schemas/code/narrative/contents, tutorial completeness, live API checks, navigator checks, catalogue reproducibility and search-memory policy passed.
• Three corrected regularization-weight plotting cells rendered separately with a live full-size Delaunay mapper.
• Strict environment-profile validation passes after moving the full-data settings from profile overrides into the scripts. Regeneration leaves notebook and catalogue content unchanged.
• Non-linear searches used sampler bypass with actual likelihood evaluations; posterior convergence was not tested.

Closes #125.

Readiness

Heart YELLOW acknowledged by the human on 2026-11-04: workspace validation not passing (0 failed, 1 timeout, cloud#41432188469: autofit_test scripts/searches/MultiStartResumeNaNCounters.py). Development shipping only; GitHub CI remains required for merge.