diff --git a/config/quality/eslint-suppressions.json b/config/quality/eslint-suppressions.json
index 6d9ef3f219..255c7eeb5e 100644
--- a/config/quality/eslint-suppressions.json
+++ b/config/quality/eslint-suppressions.json
@@ -1715,6 +1715,11 @@
       "count": 2
     }
   },
+  "tests/unit/oauth-refresh-connection-dedup-8059.test.ts": {
+    "@typescript-eslint/no-explicit-any": {
+      "count": 1
+    }
+  },
   "tests/unit/observability-fase04.test.ts": {
     "@typescript-eslint/no-explicit-any": {
       "count": 1
diff --git a/open-sse/services/accountFallback.ts b/open-sse/services/accountFallback.ts
index c10ade929f..0f43303e81 100644
--- a/open-sse/services/accountFallback.ts
+++ b/open-sse/services/accountFallback.ts
@@ -591,18 +591,7 @@ export function recordModelLockoutFailure(
   options: {
     exactCooldownMs?: number | null;
     maxCooldownMs?: number;
-    /**
-     * #6863 vs #7940: set true only when `exactCooldownMs` was parsed/verified from
-     * an actual upstream signal (Retry-After header, X-RateLimit-Reset, or a reset
-     * parsed from the error body — i.e. `usedUpstreamRetryHint`/`quotaResetHintMs`
-     * from `checkFallbackError`). A verified reset is honored exactly, even past
-     * `maxCooldownMs` — a real "Resets in 92h" must not be clamped down to minutes,
-     * or the router hammers 429 against quota that is known not to come back.
-     * Leave false/omitted for SYNTHETIC estimates (e.g. the quota_exhausted
-     * until-midnight default below, or plain exponential backoff) — those stay
-     * capped, per #7940.
-     */
-    exactCooldownVerified?: boolean;
+    exactCooldownIsUpstreamReset?: boolean;
   } = {}
 ) {
   ensureCleanupTimer();
@@ -626,19 +615,18 @@ export function recordModelLockoutFailure(
   const failureCount = withinWindow ? previous.failureCount + 1 : 1;
 
   const baseCooldownMs = getModelLockBaseCooldown(status, fallbackCooldownMs, profile);
-  // Cap exponential backoff and SYNTHETIC exact cooldowns (e.g. the daily-quota
-  // until-midnight heuristic below) against maxCooldownMs so user-configured caps
-  // are honored (#7940). A caller-VERIFIED exact cooldown (#6863 — parsed from an
-  // actual upstream Retry-After/reset signal, see `exactCooldownVerified` above)
-  // bypasses the cap instead of being clamped to a window the upstream already
-  // told us is wrong.
+  // Cap both exponential backoff and computed exact cooldowns (e.g. daily-quota
+  // until-midnight, #7940/#7980) against maxCooldownMs so user-configured caps are
+  // honored — EXCEPT an authoritative parsed upstream reset (#6863, e.g. Antigravity
+  // "Resets in 92h27m28s"), which the upstream told us to wait and must be honored
+  // exactly, never clamped down to maxCooldownMs.
   const maxCooldownMs =
     typeof options.maxCooldownMs === "number" && options.maxCooldownMs > 0
       ? options.maxCooldownMs
       : null;
   const cooldownMs =
     typeof options.exactCooldownMs === "number" && options.exactCooldownMs > 0
-      ? maxCooldownMs !== null && !options.exactCooldownVerified
+      ? maxCooldownMs !== null && !options.exactCooldownIsUpstreamReset
         ? Math.min(options.exactCooldownMs, maxCooldownMs)
         : options.exactCooldownMs
       : Math.min(
diff --git a/open-sse/services/combo.ts b/open-sse/services/combo.ts
index e76bd3e4e9..58a74cfebc 100644
--- a/open-sse/services/combo.ts
+++ b/open-sse/services/combo.ts
@@ -2425,7 +2425,10 @@ export async function handleComboChat({
                     // upstream reset (lockoutHintVerified) bypasses it.
                     exactCooldownMs: selectLockoutCooldownMs(lockoutHintMs, mlSettings),
                     maxCooldownMs: mlSettings.maxCooldownMs,
-                    exactCooldownVerified: lockoutHintVerified,
+                    // #6863: a parsed upstream quota reset is authoritative — the upstream
+                    // told us exactly when it resets, so honor it in full instead of
+                    // clamping to maxCooldownMs (which only bounds computed backoff).
+                    exactCooldownIsUpstreamReset: lockoutHintMs > mlSettings.baseCooldownMs,
                   }
                 );
                 lockoutRecorded = true;
@@ -2479,7 +2482,9 @@ export async function handleComboChat({
                   // upstream reset (lockoutHintVerified) bypasses it.
                   exactCooldownMs: selectLockoutCooldownMs(lockoutHintMs, mlSettings),
                   maxCooldownMs: mlSettings.maxCooldownMs,
-                  exactCooldownVerified: lockoutHintVerified,
+                  // #6863: an authoritative parsed upstream reset must be honored in full,
+                  // never clamped to maxCooldownMs (which only bounds computed backoff).
+                  exactCooldownIsUpstreamReset: lockoutHintMs > mlSettings.baseCooldownMs,
                 }
               );
             }
diff --git a/open-sse/utils/reasoningPlaceholder.ts b/open-sse/utils/reasoningPlaceholder.ts
index c552301cdb..915af48c92 100644
--- a/open-sse/utils/reasoningPlaceholder.ts
+++ b/open-sse/utils/reasoningPlaceholder.ts
@@ -12,18 +12,15 @@ export function isInternalReasoningPlaceholder(value: unknown): boolean {
 /**
  * Strip the internal placeholder from user-visible content. Models sometimes
  * echo the sentinel through ordinary `message.content` / `delta.content`
- * (#8081). Removes all occurrences and trims; returns "" when nothing
- * meaningful remains so callers can skip emission entirely.
+ * (#8081). Removes all occurrences; returns "" when only whitespace remains so
+ * callers can skip emission entirely.
  *
- * The trim only applies when the sentinel was actually present. This is
- * called per streaming `delta.content` chunk, not on the fully-assembled
- * message — tokenizers routinely emit sub-word tokens with a leading space
- * as part of the token (e.g. " en", " riktig"), so unconditionally trimming
- * every chunk silently ate the space between words for the (overwhelming)
- * majority of chunks that never contain the sentinel at all, producing
- * streamed text with words run together ("Bilden är en" -> "Bildenären").
+ * IMPORTANT (#5786): this runs per-delta on the streaming path, where a delta's
+ * leading/trailing spaces are meaningful (e.g. "Hello, " + "world." + " Bye.").
+ * Only collapse to "" when the placeholder WAS the whole content — never trim
+ * real content, or streamed deltas glue together with their spaces eaten.
  */
 export function stripInternalReasoningPlaceholder(value: string): string {
-  if (!value.includes(NON_ANTHROPIC_THINKING_PLACEHOLDER)) return value;
-  return value.replaceAll(NON_ANTHROPIC_THINKING_PLACEHOLDER, "").trim();
+  const stripped = value.replaceAll(NON_ANTHROPIC_THINKING_PLACEHOLDER, "");
+  return stripped.trim() === "" ? "" : stripped;
 }
diff --git a/src/sse/services/auth.ts b/src/sse/services/auth.ts
index 3eaeb7a737..0d3906acc8 100644
--- a/src/sse/services/auth.ts
+++ b/src/sse/services/auth.ts
@@ -2064,13 +2064,6 @@ export async function markAccountUnavailable(
               ? fallbackResult.cooldownMs
               : (fallbackResult.quotaResetHintMs ?? null),
           maxCooldownMs: mlSettings.maxCooldownMs,
-          // #6863 vs #7940: exactCooldownMs above is only ever set from a genuine
-          // upstream signal (Retry-After/reset header or a parsed quotaResetHintMs) —
-          // never a synthetic estimate — so it must bypass maxCooldownMs instead of
-          // being clamped down to a window the upstream already told us is wrong.
-          exactCooldownVerified:
-            fallbackResult.usedUpstreamRetryHint === true ||
-            typeof fallbackResult.quotaResetHintMs === "number",
         }
       );
       // Update last error for observability (without changing terminal status)
@@ -2140,10 +2133,6 @@ export async function markAccountUnavailable(
           exactCooldownMs:
             fallbackResult.usedUpstreamRetryHint === true ? fallbackResult.cooldownMs : null,
           maxCooldownMs: mlSettings.maxCooldownMs,
-          // #6863 vs #7940: only a genuine upstream retry hint bypasses maxCooldownMs;
-          // absent a hint, exactCooldownMs above is null and this falls through to the
-          // (still-capped) exponential-backoff branch.
-          exactCooldownVerified: fallbackResult.usedUpstreamRetryHint === true,
         }
       );
       updateProviderConnection(connectionId, {
diff --git a/tests/unit/authz/spawn-capable-prefixes-client-safe.test.ts b/tests/unit/authz/spawn-capable-prefixes-client-safe.test.ts
index dc8782b993..8b1ee022eb 100644
--- a/tests/unit/authz/spawn-capable-prefixes-client-safe.test.ts
+++ b/tests/unit/authz/spawn-capable-prefixes-client-safe.test.ts
@@ -80,7 +80,7 @@ test("SPAWN_CAPABLE_PREFIXES is defined in the server-free constants leaf with t
     "/api/skills/collect/",
     "/api/headroom/start",
     "/api/headroom/stop",
-    "/api/vnc-session", // #7892: spawns Docker containers via child_process.spawn
+    "/api/vnc-session",
   ]) {
     assert.ok(
       SPAWN_CAPABLE_PREFIXES.includes(prefix),