Stabilize diagnostic window statistics at extreme scale

[FlorianPfaff]

Bug

worst_time_windows(...) computed RMSE as sqrt(mean(errors**2)). For large but finite errors, the squaring step can overflow even when the mathematically correct RMSE is finite. The same window could also report non-finite MAE or percentile values because their intermediate reductions/interpolation operate at the original extreme scale.

A concrete finite example is a window with errors [-1e308, 1e308]. Its correct statistics are finite (RMSE = 1e308, MAE = 1e308, p95 = 9e307), while the previous implementation overflows under strict NumPy floating-point handling.

Fix

Scale each window's finite errors by their maximum absolute magnitude before computing RMSE, MAE, and the 95th percentile, then restore the original scale. The normalized calculations stay in [-1, 1], avoiding overflow while preserving ordinary-scale results. Zero-error windows retain an explicit zero fast path.

Regression coverage

Add a focused regression using [-1e308, 1e308] under np.errstate(over="raise", invalid="raise"). It verifies all three statistics remain finite and match their expected values.

Validation

• branch is based directly on current main (f196ec41d1d8681e279936db9f1a0a3abfa276e7);
• branch is 2 commits ahead / 0 behind;
• final diff is limited to 16 changed lines in diagnostic_summaries.py plus one 22-line regression module;
• independently reproduced the old overflow and verified the scale-normalized calculation yields finite 1e308, 1e308, and 9e307 results;
• GitHub Actions is authoritative for the full repository test, lint, packaging, documentation, and integration matrix.

[github-actions]

✅MegaLinter analysis: Success

Descriptor	Linter	Files	Fixed	Errors	Warnings	Elapsed time
✅ COPYPASTE	jscpd	yes		no	no	24.56s
✅ JSON	prettier	7	0	0	0	1.27s
✅ JSON	v8r	7		0	0	4.22s
✅ MARKDOWN	markdownlint	68	0	0	0	1.95s
✅ MARKDOWN	markdown-table-formatter	68	0	0	0	0.62s
✅ PYTHON	black	1867	87	0	0	89.38s
✅ PYTHON	isort	1867	141	0	0	2.75s
✅ REPOSITORY	betterleaks	yes		no	no	2.28s
✅ REPOSITORY	checkov	yes		no	no	53.47s
✅ REPOSITORY	gitleaks	yes		no	no	17.6s
✅ REPOSITORY	git_diff	yes		no	no	0.24s
✅ REPOSITORY	secretlint	yes		no	no	80.95s
✅ REPOSITORY	syft	yes		no	no	5.49s
✅ REPOSITORY	trivy-sbom	yes		no	no	6.64s
✅ REPOSITORY	trufflehog	yes		no	no	30.6s
✅ YAML	prettier	11	0	0	0	0.8s
✅ YAML	v8r	11		0	0	11.77s
✅ YAML	yamllint	11		0	0	0.58s

Notices

📣 MegaLinter 9.5.0 is out! Discover the new features and security recommendations in the release announcement. (Skip this info by defining SECURITY_SUGGESTIONS: false)

See detailed reports in MegaLinter artifacts

Your project could benefit from a custom flavor, which would allow you to run only the linters you need, and thus improve runtime performances. (Skip this info by defining FLAVOR_SUGGESTIONS: false)

• Documentation: Custom Flavors
• Command: npx mega-linter-runner@9.6.0 --custom-flavor-setup --custom-flavor-linters PYTHON_BLACK,PYTHON_ISORT,COPYPASTE_JSCPD,JSON_V8R,JSON_PRETTIER,MARKDOWN_MARKDOWNLINT,MARKDOWN_MARKDOWN_TABLE_FORMATTER,REPOSITORY_CHECKOV,REPOSITORY_GIT_DIFF,REPOSITORY_GITLEAKS,REPOSITORY_BETTERLEAKS,REPOSITORY_SECRETLINT,REPOSITORY_SYFT,REPOSITORY_TRIVY_SBOM,REPOSITORY_TRUFFLEHOG,YAML_PRETTIER,YAML_YAMLLINT,YAML_V8R

Show us your support by starring ⭐ the repository