fix(analytics): retire the research view and profile update event types

[quntao-z]

Intent

Retire the two analytics event types that nothing emits any more, now that their stored history was deleted from Development and Production on 2026-10-11 by the owner's decision, so every stored event type has a live emitter and no dead type can feed a dashboard total. Record the decision in docs/decisions.md as superseding the 2026-08-29 keep-for-history rule. PR body rule: describe the regression tests by behaviour only and never quote a test fixture string, slug, or name from the diff, because the guard scans the body and a quoted fixture reads as a real person. Do not quote dashboard tile names in capitals. Link the issue with Closes #4992 and state that the data operation already ran on Development and Production.

What Changed

• Removes the research_view and profile_update members from AnalyticsEventType. Nothing emits either type any more. Their reads are gone from analyticsService: the research engagement totals, the search-quality attribution window, and the per-user profileUpdates count, which is also dropped from AnalyticsUserSummary and the client AnalyticsUserActivityRow.
• Extends the per-user activity regression test so it also asserts that the two retired types are never counted. The client analytics page tests and the search-attribution equivalence test no longer use the removed column or enum member.
• Adds a 2026-10-11 entry to docs/decisions.md that supersedes the 2026-08-29 keep-for-history rule, and marks the older entry as superseded. It also updates the analytics docs and DEVELOPER_GUIDE.md so they match.

The data operation already ran on Development and Production: every stored row of the two types was deleted, with a local backup. Beta held none. Every stored event type now has a live emitter.

Closes #4992

Risk Assessment

✅ Low: The change only removes two enum members and the dashboard reads, column and attribution entries that depended on them, and no code anywhere in the repo still references either type; the batch route already rejects unknown types, and the decisions entry correctly supersedes the 2026-08-29 rule.

Testing

I ran the real API on a throwaway local MongoDB seeded with synthetic legacy rows, then ran the base commit against the same data. Base counted the 3 legacy research-view rows in the research totals, gave one user 5 events, and listed a user who had only retired rows. Target showed only the live research_profile_open type, gave that user 2 events, and dropped the retired-only user. It also returned no profileUpdates field anywhere. The research batch endpoint treats both retired types as unrecognized. The focused analytics tests pass. No screenshot was taken: the client side only removes a field from a TypeScript type and changes no rendered surface. The emitter check and the decisions-doc check were done by reading source, not live, so they are marked untested. All temporary DB, server and script files were torn down.

• Live validation: ✅ go - 3 of 5 scenarios driven live against the product

Scenario	Result	Live	Evidence
Admin opens the analytics overview while legacy research_view rows are still stored, and the research tile counts no research_view	✅ pass	live	analytics-overview.json: research.byEventType is only [research_profile_open total 1]. baseline-analytics-overview.json shows research_view total 3.
Admin opens the per-user activity list, and leftover profile_update/research_view rows neither inflate totals nor appear as a column	✅ pass	live	analytics-users.json: the synthetic user has totalEvents 2 (base said 5), there is no profileUpdates field, and the user with only retired rows is not listed (base listed them)
Adversarial: a client posts a retired event type to the research batch endpoint, and it is not stored	✅ pass	live	research-batch-ingest.txt plus the server log: research_view and profile_update come back accepted 0 and are tallied as 'unrecognized'. The research_save control is recognized.
Every remaining event type still has a live emitter, logout included, and nothing in source still emits the retired types	⏸️ untested	no	The prior payload did not establish a live result: this was checked only by grepping source (no emitter of research_view or profile_update found, and LOGOUT is emitted at server/src/passport.ts:819),…
The decision is recorded in docs/decisions.md as superseding the 2026-08-29 keep-for-history rule	⏸️ untested	no	The prior payload did not establish a live result: this is a docs-only requirement with no runtime surface, and it was checked only by reading the docs/decisions.md diff.

Evidence: Target /api/analytics overview (retired types absent)

{"visitors":{"lifetime":{"total":1,"byType":[{"userType":"undergraduate","count":1}]},"last7Days":{"total":1,"byType":[{"userType":"undergraduate","count":1}]},"today":{"total":1,"byType":[{"userType":"undergraduate","count":1}]},"loginFrequency":{"totalLogins":1,"loginsLast7Days":1,"loginsToday":1}},"engagement":{"search":{"_id":null,"totalSearches":1,"searchesLast7Days":1,"searchesToday":1},"topSearchQueries":[],"userActivity":{"_id":null,"activeUsers":1,"avgEventsPerUser":3},"mostActiveUsers":[{"eventCount":2,"userId":"synthstu01","userType":"undergraduate"}]},"research":{"byEventType":[{"total":1,"last7Days":1,"today":1,"eventType":"research_profile_open"}],"byEntityType":[],"byUserType":[{"userType":"undergraduate","count":1}],"topEntities":[]},"users":{"overview":{"_id":null,"total":1,"confirmed":1},"byType":[],"newUsersLast7Days":1,"newUsersToday":1,"newUsersTodayByType":[]},"researchEntities":{"overview":{"active":0,"total":0},"byType":[],"byVisibilityTier":[],"freshness":{"observedLast7Days":0,"observedLast30Days":0,"neverObserved":0,"staleOver90Days":0},"scholarly":{"withRecentGrants":0}},"timestamp":"2026-10-11T20:54:07.112Z","timeZone":"America/New_York"}

Evidence: Target /api/analytics/users (no profileUpdates, retired-only user absent)

{"users":[{"totalEvents":1,"lastEventAt":"2026-10-11T20:54:07.109Z","logins":0,"researchViews":0,"fellowshipViews":0,"netid":"devadmin","userType":"admin","firstSeen":"2026-10-11T20:54:07.066Z","lastActive":"2026-10-11T20:54:07.146Z","lastLogin":"2026-10-11T20:54:07.146Z"},{"totalEvents":2,"lastEventAt":"2026-10-11T20:54:01.236Z","logins":1,"researchViews":1,"fellowshipViews":0,"netid":"synthstu01","userType":"undergraduate","firstSeen":"2026-10-11T20:54:01.236Z","lastActive":"2026-10-11T20:54:01.236Z"}],"total":2,"limit":50,"offset":0}

Evidence: Baseline /api/analytics overview (research_view total 3)

{"visitors":{"lifetime":{"total":2,"byType":[{"userType":"undergraduate","count":2}]},"last7Days":{"total":2,"byType":[{"userType":"undergraduate","count":2}]},"today":{"total":2,"byType":[{"userType":"undergraduate","count":2}]},"loginFrequency":{"totalLogins":1,"loginsLast7Days":1,"loginsToday":1}},"engagement":{"search":{"_id":null,"totalSearches":1,"searchesLast7Days":1,"searchesToday":1},"topSearchQueries":[],"userActivity":{"_id":null,"activeUsers":2,"avgEventsPerUser":4},"mostActiveUsers":[{"eventCount":5,"userId":"synthstu01","userType":"undergraduate"},{"eventCount":2,"userId":"synthstu02","userType":"undergraduate"}]},"research":{"byEventType":[{"total":3,"last7Days":3,"today":3,"eventType":"research_view"},{"total":1,"last7Days":1,"today":1,"eventType":"research_profile_open"}],"byEntityType":[],"byUserType":[{"userType":"undergraduate","count":4}],"topEntities":[]},"users":{"overview":{"_id":null,"total":1,"confirmed":1},"byType":[],"newUsersLast7Days":1,"newUsersToday":1,"newUsersTodayByType":[]},"researchEntities":{"overview":{"active":0,"total":0},"byType":[],"byVisibilityTier":[],"freshness":{"observedLast7Days":0,"observedLast30Days":0,"neverObserved":0,"staleOver90Days":0},"scholarly":{"withRecentGrants":0}},"timestamp":"2026-10-11T20:54:51.289Z","timeZone":"America/New_York"}

Evidence: Baseline /api/analytics/users (inflated totals, profileUpdates column)

{"users":[{"totalEvents":1,"lastEventAt":"2026-10-11T20:54:07.109Z","logins":0,"researchViews":0,"fellowshipViews":0,"profileUpdates":0,"netid":"devadmin","userType":"admin","firstSeen":"2026-10-11T20:54:07.066Z","lastActive":"2026-10-11T20:54:51.324Z","lastLogin":"2026-10-11T20:54:51.324Z"},{"totalEvents":5,"lastEventAt":"2026-10-11T20:54:01.236Z","logins":1,"researchViews":1,"fellowshipViews":0,"profileUpdates":1,"netid":"synthstu01","userType":"undergraduate","firstSeen":"2026-10-11T20:54:01.236Z","lastActive":"2026-10-11T20:54:01.236Z"},{"totalEvents":2,"lastEventAt":"2026-10-11T20:54:01.236Z","logins":0,"researchViews":0,"fellowshipViews":0,"profileUpdates":1,"netid":"synthstu02","userType":"undergraduate","firstSeen":"2026-10-11T20:54:01.236Z","lastActive":"2026-10-11T20:54:01.236Z"}],"total":3,"limit":50,"offset":0}

Evidence: Research batch ingest of retired types

research_view -> accepted 0 (rejected as unrecognized); profile_update -> accepted 0 (unrecognized); research_save control -> recognized, rejected only because the synthetic entity does not exist

== research_view
{"accepted":0,"sent":1} HTTP 202
== profile_update
{"accepted":0,"sent":1} HTTP 202
== research_save
{"accepted":0,"sent":1} HTTP 202

Pipeline

Updates from git push no-mistakes

✅ **intent** - passed

✅ No issues found.

✅ **Rebase** - passed

✅ No issues found.

✅ **Review** - passed

✅ No issues found.

✅ **Test** - passed

✅ No issues found.

• Live validation: ✅ go - 3 of 5 scenarios driven live against the product

Scenario	Result	Live	Evidence
Admin opens the analytics overview while legacy research_view rows are still stored, and the research tile counts no research_view	✅ pass	live	analytics-overview.json: research.byEventType is only [research_profile_open total 1]. baseline-analytics-overview.json shows research_view total 3.
Admin opens the per-user activity list, and leftover profile_update/research_view rows neither inflate totals nor appear as a column	✅ pass	live	analytics-users.json: the synthetic user has totalEvents 2 (base said 5), there is no profileUpdates field, and the user with only retired rows is not listed (base listed them)
Adversarial: a client posts a retired event type to the research batch endpoint, and it is not stored	✅ pass	live	research-batch-ingest.txt plus the server log: research_view and profile_update come back accepted 0 and are tallied as 'unrecognized'. The research_save control is recognized.
Every remaining event type still has a live emitter, logout included, and nothing in source still emits the retired types	⏸️ untested	no	The prior payload did not establish a live result: this was checked only by grepping source (no emitter of research_view or profile_update found, and LOGOUT is emitted at server/src/passport.ts:819),…
The decision is recorded in docs/decisions.md as superseding the 2026-08-29 keep-for-history rule	⏸️ untested	no	The prior payload did not establish a live result: this is a docs-only requirement with no runtime surface, and it was checked only by reading the docs/decisions.md diff.

• Booted the real server at the target commit (npx tsx src/index.ts, NODE_ENV=development, LOCAL_AUTH_BYPASS as admin) against a throwaway mongod on 127.0.0.1:27991 with its own /tmp dbpath
• Inserted synthetic analytics_events: 3 research_view, 2 profile_update, and research_profile_open/search/login rows. One synthetic user had only retired rows.
• curl GET /api/analytics and GET /api/analytics/users as admin, on the target server (port 4992)
• The same two calls on the base commit 5790e4d7c, booted from a git archive copy on port 4993 against the same seeded DB
• curl POST /api/analytics/research/batch with research_view, profile_update, and research_save as a control, then read the server's rejection log
• grep of server/src and client/src for any remaining emitter of the retired types, and a check that LOGOUT still has an emitter (passport.ts)
• npx vitest run src/services/__tests__/analyticsService.test.ts src/services/__tests__/searchQualityAttribution.equivalence.test.ts
• Read the docs/decisions.md diff: a new 2026-10-11 entry, plus a note on the 2026-08-29 entry marking it superseded

✅ **Document** - passed

✅ No issues found.

⚠️ **Lint** - 1 info

• ℹ️ ESLint could not run because this worktree has no installed node_modules (no client, server, or root eslint binary). Prettier 3.9.9 passed on all changed .ts/.tsx files. The code diff only removes enum members, fields, and array entries, so new lint errors are unlikely.

✅ **Push** - passed

✅ No issues found.

[github-actions]

Approved automatically: the author holds the Admin repository role (#4512).