diff --git a/lib/github-api.js b/lib/github-api.js
index 8f1b2cc..148fd9e 100644
--- a/lib/github-api.js
+++ b/lib/github-api.js
@@ -48,9 +48,9 @@ async function checkRateLimit() {
 // Wait for rate limit to reset using polling
 async function waitForRateLimitReset(targetResetTime) {
   const POLL_INTERVAL = 60000; // 60 seconds
-  // In CI: use low threshold for incremental progress within timeout constraints
+  // In CI: use moderate threshold to balance efficiency and timeout constraints
   // Locally: use high threshold for better user experience (fewer interruptions)
-  const MIN_REMAINING_TO_CONTINUE = process.env.CI === 'true' ? 100 : 4500;
+  const MIN_REMAINING_TO_CONTINUE = process.env.CI === 'true' ? 1000 : 4500;
   const MAX_POLLS = 120; // Max 120 polls = 2 hours
 
   const estimatedWaitMinutes = Math.ceil(Math.max(0, targetResetTime - Date.now()) / 60000);
@@ -121,21 +121,24 @@ async function rateLimitedRequest(url, options = {}) {
     await new Promise(resolve => setTimeout(resolve, RATE_LIMIT_DELAY - timeSinceLastRequest));
   }
 
-  // Check rate limit proactively every 100 requests
-  // BUT skip proactive checks for 10 minutes after recovering from rate limit
+  // Check rate limit proactively
+  // In CI: check frequently with no grace period to maximize efficiency
+  // Locally: use grace period to reduce API overhead after recovery
+  const isCI = process.env.CI === 'true';
   const timeSinceRecovery = Date.now() - lastRateLimitRecoveryTime;
-  const RECOVERY_GRACE_PERIOD = 10 * 60 * 1000; // 10 minutes
+  const RECOVERY_GRACE_PERIOD = isCI ? 0 : (10 * 60 * 1000); // No grace period in CI, 10min locally
+  const CHECK_FREQUENCY = isCI ? 0.10 : 0.01; // 10% in CI, 1% locally
+  const LOW_THRESHOLD = isCI ? 500 : 200; // Wait at 500 in CI for full batches, 200 locally
 
-  if (timeSinceRecovery > RECOVERY_GRACE_PERIOD && Math.random() < 0.01) { // 1% chance
+  if (timeSinceRecovery > RECOVERY_GRACE_PERIOD && Math.random() < CHECK_FREQUENCY) {
     const rateLimitStatus = await checkRateLimit();
-    if (rateLimitStatus && rateLimitStatus.remaining < 200) {
+    if (rateLimitStatus && rateLimitStatus.remaining < LOW_THRESHOLD) {
       console.log();
       console.log(chalk.yellow(`⚠️  Rate limit getting low: ${rateLimitStatus.remaining}/${rateLimitStatus.limit} remaining`));
       console.log(chalk.yellow(`   Proactively waiting for rate limit to reset...`));
 
-      const isCI = process.env.CI === 'true';
       if (isCI) {
-        console.log(chalk.cyan('🤖 CI mode: starting polling to wait for reset...'));
+        console.log(chalk.cyan('🤖 CI mode: waiting for full reset to maximize batch efficiency...'));
         await waitForRateLimitReset(rateLimitStatus.reset);
       }
     }