Stop Community Notes from burying organic For You rank

[Pitchfork-and-Torch]

Bug

Community Notes were ads MediumRisk.

NSFA_COMMUNITY_NOTE sat in MEDIUM_RISK_LABELS and MEDIUM_RISK_LABELS_V2. AdsBrandSafetyVfHydrator writes that verdict onto the shared post. has_avoid treats MediumRisk as ads-avoid. Default For You blenders (partition_organic, time_gap) then pull avoid posts out of Phoenix order and dump them into filler after the ad sandwich.

A reply or quote inherits the worst ancestor/quoted verdict. A note on the parent buried the whole conversation when ads placed.

This is not #127 (missing verdict / nsfw_author_ads leaking MediumRisk). #127 keeps real MediumRisk off ad neighbors. A crowd note is not brand-unsafe inventory. VF does not drop NSFA_COMMUNITY_NOTE. Phoenix does not score it. Ads taxonomy was the only rank writer.

Fix

Remove NSFA_COMMUNITY_NOTE from v1 and v2 MediumRisk. A Grok-safe noted post stays Safe. Real NSFW/NSFA and unscored posts stay MediumRisk.

Proof

• Entry: NSFA_COMMUNITY_NOTE on the tweet, quote, or ancestor
• Sink: compute_verdict / compute_verdict_v2 -> has_avoid -> partition_organic / time_gap filler
• Break: a crowd note was classed with GROK_NSFA and MALICIOUS_URL; the blender then reordered organics
• Viewer effect: noted post (or its reply) drops below a lower-scored Safe post whenever ads run
• Twin: GROK_SFA-only stays Safe and keeps rank; GROK_SFA+GROK_NSFA stays MediumRisk

Tests

• community_note_is_not_ads_medium_risk (v1 and v2)
• community_note_does_not_mask_real_medium_or_unscored
• community_note_keeps_safe_verdict
• community_note_on_quoted_or_ancestor_does_not_escalate
• medium_risk_avoid_buries_higher_scored_organic (sink lock)
• v2_mirrors_v1_across_tier_matrix includes the note pair
• unit tests added; cargo test cannot run in the public dump

[Pitchfork-and-Torch]

Parking this intern-stack PR per operator GitHub cleanup (2026-09-08). Hunt notes remain local. Not a reject of the class.