[hyphaltip]

The v0.9.0 release PR on the path the curator chose on 2026-11-12: the 372 genomes that moved under the tblastn outlier policy are taken from reruns with it OFF, and the GCA_052323795.1 outer-span extension (+16.7 kb, #302) is approved. Not merged, not tagged by me.

What

• Results: results/2026-11-09_v090_{ascomycota,basidiomycota,mucoromycotina} (tables, lists, manifests, reports_all.tar.zst). The Ascomycota tarball holds the 372 SEG-off reruns instead of the 372 SEG-on reports: 19,343 reports, one per genome, no duplicates; Basidiomycota 3,263; Mucoromycotina 973.
• scripts/build_site_data.py: v090 in RELEASES and DEFAULT_RELEASE = "v090" (v081 stays as the rollback); test updated.
• Site data regenerated on the HPCC; README, About and landing text describe v0.9.0.

Checks

• Full suite: 1,989 passed, 20 skipped.
• Site build from these folders on the HPCC; counts agree with the tables (17,618 / 20,844 Ascomycota, 2,995 / 5,994 Basidiomycota).
• Build-only Pages run 43200546755 succeeded (it does not deploy); the landing and results pages rendered from its artifact with no console errors.

For the curator

• Merge order: this PR, then CHANGELOG: v0.9.0 definitive-run counts (SEG-off replacement chosen) #334 (CHANGELOG counts; both touch different files), then tag. The Pages deploy runs on the v* tag.
• Size: about 71 MB of tarballs and tables (v0.8.1 was about 31 MB).
• The PRs merged after the frozen run (Scoring: genes_found lists optional slot genes in roster order (reproducible report order) #335 gene order, #324: record an empty taxonomy lookup (routing_error); panel warm-up counts it and can stop the job #339 taxonomy diagnostics) are not in the run's code stamp (7d73be5); they change no call.

🤖 Generated with Claude Code

[hyphaltip]

Site build check (HPCC, job 32038515, --release v090 into a scratch folder; the repo's web/ data and the default release are untouched): build_site_data.py and build_locus_shards.py both complete from these folders. Counts agree with the tables: bfd_asco 17,618 genomes / 20,844 loci, bfd_basidio 2,995 / 5,994, 23,664 genomes in all. Still not done: the default release switch, README/About/landing text, and the committed web/intro/data and shards, which follow the curator's choice.