perf(contests): swap aggregate_user.score filter for karma-reported authors

[dylanjeffers]

Problem

/v1/events/remix-contests?limit=12&offset=0&status=all is timing out / hanging the contests page in production. Measured against api.audius.co:

Call	First (cold)	Subsequent (warm)
status=all	24.7s	0.12s
status=ended	24.0s	0.12s
status=active	1.96s	0.08s

Root cause

The shadowban filter introduced in #803 added an aggregate_user.score < 0 CTE on the contest query. aggregate_user has one row per user (millions of rows) and no index covering score, so the CTE runs a full sequential scan of aggregate_user on every cold call.

WITH low_abuse_score AS (SELECT user_id FROM aggregate_user WHERE score < 0)
...
WHERE e.user_id NOT IN (SELECT user_id FROM low_abuse_score)

Approach — align with the existing comment-visibility filter

Rather than indexing around aggregate_user.score, this PR switches the contest filter to the same community-karma signal that already governs comment visibility on v1_track_comments / v1_event_comments. A comment is hidden when reporters' summed follower_count crosses karmaCommentCountThreshold (the high_karma_reporters CTE). This PR lifts that signal from comment_id up to user_id: if a host has authored any comment that crossed the threshold, their contests are hidden too.

New CTEs on the contest query:

high_karma_reporters AS (
    SELECT comment_reports.comment_id
    FROM comment_reports
    JOIN aggregate_user ON comment_reports.user_id = aggregate_user.user_id
    WHERE comment_reports.is_delete = false
    GROUP BY comment_reports.comment_id
    HAVING SUM(aggregate_user.follower_count) >= @karmaCommentCountThreshold
),
karma_reported_authors AS (
    SELECT DISTINCT comments.user_id
    FROM comments
    JOIN high_karma_reporters USING (comment_id)
)

Filter:

WHERE e.user_id NOT IN (SELECT user_id FROM karma_reported_authors)
  AND e.user_id NOT IN (SELECT muted_user_id FROM muted_by_karma)

muted_by_karma is unchanged — still in the filter list. The AAO score remains the moderation lane for the comment endpoints themselves (v1_track_comments, v1_event_comments, v1_fan_club_feed, v1_track_comment_count are all untouched).

Why this is faster

The new CTE walks comment_reports (a small table, indexed on comment_id) joined with aggregate_user on the reporter id, plus a hash-join on comments for the author lookup. The cost is bounded by report volume — no sequential scan over millions of aggregate_user rows.

Supersedes #813

#813 proposed a partial index on aggregate_user (user_id) WHERE score < 0. That works but keeps the AAO score as the moderation lane for contests; this PR routes contests through the same karma-reports signal as comments, which is the preferred semantic. Closing #813.

Test plan

• go build ./api/... clean
• go vet ./api/... clean
• TestRemixContestsExcludesShadowbannedHosts updated to seed comment_reports + aggregate_user.follower_count bump and assert the karma-reported host is excluded
• CI test suite green
• Time /v1/events/remix-contests?status=all cold on staging — expect sub-second
• Confirm a host with karma-reported comments is excluded
• Confirm karma-muted hosts are still excluded (unchanged)
• Comment endpoints (track / event / fan-club) unchanged — low_abuse_score, high_karma_reporters, muted_by_karma, deactivated_users all still in place

🤖 Generated with Claude Code