fix(FN-703): fix overly broad usage limit pattern and add transient error retry

- Narrow 'insufficient' usage-limit regex to require quota/credit/balance/fund context
- Add transient error detection in executor and triage to retry instead of failing
- Improve scheduler pause/resume logging with actionable messages and resume confirmation
- Add SSE active connection counter for observability
- Add duplicate usage-limit pause suppression logging

packages/dashboard/src/sse.ts

@@ -1,6 +1,13 @@
 import type { Request, Response } from "express";
 import type { TaskStore } from "@fusion/core";
 
+let activeConnections = 0;
+
+/** Returns the current number of active SSE connections. */
+export function getActiveSSEConnections(): number {
+  return activeConnections;
+}
+
 export function createSSE(store: TaskStore) {
   return (_req: Request, res: Response) => {
     res.setHeader("Content-Type", "text/event-stream");
@@ -9,6 +16,8 @@ export function createSSE(store: TaskStore) {
     res.setHeader("X-Accel-Buffering", "no");
     res.flushHeaders();
 
+    activeConnections++;
+
     // Send initial heartbeat
     res.write(": connected\n\n");
 
@@ -40,6 +49,7 @@ export function createSSE(store: TaskStore) {
     }, 30_000);
 
     _req.on("close", () => {
+      activeConnections--;
       clearInterval(heartbeat);
       store.off("task:created", onCreated);
       store.off("task:moved", onMoved);

packages/engine/src/executor.ts

@@ -13,6 +13,7 @@ import type { WorktreePool } from "./worktree-pool.js";
 import { AgentLogger } from "./agent-logger.js";
 import { executorLog, reviewerLog } from "./logger.js";
 import { isUsageLimitError, checkSessionError, type UsageLimitPauser } from "./usage-limit-detector.js";
+import { isTransientError } from "./transient-error-detector.js";
 import type { StuckTaskDetector } from "./stuck-task-detector.js";
 
 // Re-export for backward compatibility (tests import from executor.ts)
@@ -688,6 +689,12 @@ export class TaskExecutor {
         // Check if the error is a usage-limit error and trigger global pause
         if (this.options.usageLimitPauser && isUsageLimitError(err.message)) {
           await this.options.usageLimitPauser.onUsageLimitHit("executor", task.id, err.message);
+        } else if (isTransientError(err.message)) {
+          // Transient network/infrastructure error — retry instead of failing
+          executorLog.warn(`⚡ ${task.id} transient error — moving to todo for retry: ${err.message}`);
+          await this.store.logEntry(task.id, `Transient error (will retry): ${err.message}`);
+          await this.store.moveTask(task.id, "todo");
+          return;
         }
         executorLog.error(`✗ ${task.id} execution failed:`, err.message);
         await this.store.logEntry(task.id, `Execution failed: ${err.message}`);

packages/engine/src/scheduler.ts

@@ -331,21 +331,27 @@ export class Scheduler {
       // Global pause (hard stop): halt all scheduling activity
       if (settings.globalPause) {
         if (!this.wasGlobalPaused) {
-          schedulerLog.log("Global pause active — scheduling halted");
+          schedulerLog.warn("⚠ Global pause active — scheduling halted. To resume: set globalPause to false in settings.");
           this.wasGlobalPaused = true;
         }
         return;
       }
+      if (this.wasGlobalPaused) {
+        schedulerLog.log("Global pause cleared — scheduling resumed");
+      }
       this.wasGlobalPaused = false;
 
       // Engine paused (soft pause): halt new work dispatch, but let agents finish
       if (settings.enginePaused) {
         if (!this.wasEnginePaused) {
-          schedulerLog.log("Engine paused — scheduling halted (in-flight agents continue)");
+          schedulerLog.warn("⚠ Engine paused — scheduling halted (in-flight agents continue). To resume: set enginePaused to false.");
           this.wasEnginePaused = true;
         }
         return;
       }
+      if (this.wasEnginePaused) {
+        schedulerLog.log("Engine pause cleared — scheduling resumed");
+      }
       this.wasEnginePaused = false;
 
       // Count only in-progress tasks toward the worktree limit.

packages/engine/src/triage.ts

@@ -21,6 +21,7 @@ import {
   checkSessionError,
   type UsageLimitPauser,
 } from "./usage-limit-detector.js";
+import { isTransientError } from "./transient-error-detector.js";
 
 export const TRIAGE_SYSTEM_PROMPT = `You are a task specification agent for "kb", an AI-orchestrated task board.
 
@@ -664,6 +665,14 @@ export class TriageProcessor {
             task.id,
             err.message,
           );
+        } else if (isTransientError(err.message)) {
+          // Transient network/infrastructure error — don't mark as failed, allow retry
+          triageLog.warn(`⚡ ${task.id} transient error during triage — will retry: ${err.message}`);
+          await this.store.logEntry(task.id, `Transient error during specification (will retry): ${err.message}`).catch(() => {});
+          // Restore status so triage picks it up again on next pass
+          const restoreStatus = task.status === "needs-respecify" ? "needs-respecify" : undefined;
+          await this.store.updateTask(task.id, { status: restoreStatus }).catch(() => {});
+          return;
         }
         // For re-specification, restore needs-respecify status so it can be retried
         const restoreStatus = task.status === "needs-respecify" ? "needs-respecify" : undefined;

packages/engine/src/usage-limit-detector.ts

@@ -27,7 +27,7 @@ const USAGE_LIMIT_PATTERNS: RegExp[] = [
   /quota/i,
   /billing/i,
   /\bcredit/i,
-  /insufficient/i,
+  /insufficient.*(quota|credit|balance|fund)/i,
 ];
 
 /**
@@ -88,6 +88,7 @@ export class UsageLimitPauser {
       const settings = await this.store.getSettings();
       if (settings.globalPause) {
         // Still paused — no need to trigger again
+        log.log(`Global pause already active — ignoring duplicate from ${agentType}/${taskId}`);
         return;
       }
       // External reset detected — allow re-triggering
@@ -97,6 +98,7 @@ export class UsageLimitPauser {
     this.paused = true;
 
     log.warn(`${agentType} hit usage limit on ${taskId}: ${errorMessage}`);
+    log.warn(`Matched pattern in error: "${errorMessage.slice(0, 200)}"`);
 
     // Log the triggering error on the task
     await this.store.logEntry(