Problem

An anonymous dashboard load dispatches 5 summarize-article round-trips that cannot produce a summary, then logs [Summarization] All providers failed — a warning that reads like a provider outage but is the designed anon path.

Measured on a live anonymous www.worldmonitor.app/dashboard session:

performance.getEntriesByType('resource').filter(e => e.name.includes('summarize'))
// 5 requests — [status, startMs, durationMs]
[[200,1404,447],[200,1404,487],[200,1404,558],[200,1405,516],[200,1405,338]]

All five return 200 — the server gate added for #4675/#4913 is working correctly and declines gracefully (resp.fallback / SUMMARIZE_STATUS_SKIPPED). This is not a 401/403 flood; #4913 and #5207 are genuinely fixed. The residual cost is 5 concurrent round-trips of ~340–560 ms each, fired in a ~1 ms burst, that are guaranteed no-ops for this user class.

Root cause — #4913 fix direction 1 never landed

#4913's proposed fix had two parts. Part 2 (server-side graceful gate) shipped. Part 1 did not:

1. Enqueue gate: check panel-gating.hasPremiumAccess() before dispatching any API-provider summarize attempt; anon/free goes straight to the browser T5 fallback (the designed anon path) with zero network calls.

src/services/summarization.ts contains no reference to hasPremiumAccess or panel-gating (origin/main c45acd49c):

$ grep -n "hasPremiumAccess\|panel-gating" src/services/summarization.ts
(no matches)

Its only pre-dispatch check is isFeatureAvailable(providerDef.featureId) (src/services/summarization.ts:338), which gates on provider/feature availability, not on the caller's entitlement. So the chain still fans out across API_PROVIDERS for users who can never receive a summary.

Then src/services/summarization.ts:320 reports the outcome as a failure:

console.warn('[Summarization] All providers failed');

Two call sites reach it per load — trending-keywords and panels-intel updateInsights:

at c0 (main-BXxARTx3.js:2:129385)
at l0 → G (trending-keywords-pwCxGERt.js:1:7357)
at l0 → updateFromClient (panels-intel-BwxhCj3q.js:616:41) → updateInsights → loadNews

Why this is worth fixing despite being small

• The log is actively misleading. "All providers failed" is indistinguishable from a real provider outage in console triage. It cost this investigation two wrong hypotheses (first a 403 outage, then a 401 flood) before Resource Timing showed 200s. Anyone triaging prod console noise will hit the same trap.
• Same latent pattern as fix(summarization): anon dashboards dispatch premium summarize-article RPCs — 401 flood + 3-provider fan-out since #4687 (client caller never swept) #4913 / fix(classify): stop free/anon users flooding the premium-gated classify-event endpoint (#4865) #4869 / fix(health-freshness): dashboard calls gated /api/health anonymously — 401 every 60s per tab + seed-health freshness silently dead since #4715 #4902 — server gated, client caller never swept. Here it degraded from "flood" to "polite no-op", but the missing enqueue gate is the same gap and will re-cost whenever the server contract changes.
• 5 concurrent requests in the same window as the news wave (perf(pwa): service worker registers on the boot path — 882KB High-priority precache wave lands inside the paint window #5372's neighbours) on a page already issuing 211+ requests.

Proposed fix

• Add the enqueue gate from fix(summarization): anon dashboards dispatch premium summarize-article RPCs — 401 flood + 3-provider fan-out since #4687 (client caller never swept) #4913 part 1: check entitlement before dispatching any API-provider summarize attempt; anon/free goes straight to browser T5 with zero network calls.
• Distinguish "declined by design" from "genuinely failed" at src/services/summarization.ts:320 — a skipped/gated outcome should log at debug/info (or not at all), reserving warn for real provider failure. Same for the [BETA] sibling at :301.

Related

• fix(summarization): anon dashboards dispatch premium summarize-article RPCs — 401 flood + 3-provider fan-out since #4687 (client caller never swept) #4913 (closed) — fix direction 1 (enqueue gate) never implemented; part 2 shipped and works.
• bug(auth): cookieless anon browsers flood auth-required RPCs with 401s (~10k/3d on summarize alone) — client never detects a dead wms_ session #5207, Security: anonymous summarize-article requests can drive LLM provider spend #4675 (closed) — earlier 401-flood / spend variants of this path. Both confirmed fixed: statuses are 200, not 401.
• bug(frontend): free users flood classify-event with ~95k tier_403/day since #4779 premium-gated it — client callsite bypasses premiumFetch gating, no 403 backoff #4865, fix(classify): stop free/anon users flooding the premium-gated classify-event endpoint (#4865) #4869, fix(health-freshness): dashboard calls gated /api/health anonymously — 401 every 60s per tab + seed-health freshness silently dead since #4715 #4902 — same server-gate-without-client-sweep family.

Acceptance

• Anonymous dashboard load issues zero summarize-article requests; verify with performance.getEntriesByType('resource').filter(e => e.name.includes('summarize')).length === 0.
• Signed-in premium path still summarizes normally (regression check — do not gate paying users out).
• [Summarization] All providers failed no longer appears on a healthy anon load; a real provider failure still surfaces at warn.