docs: Update CI polling guidance to use 30-second intervals

Update documentation to explicitly recommend 30-second polling intervals
when waiting for CI, rather than 180 seconds. This matches the existing
behavior in bin/ci-rerun-failures script.

Added new "Polling CI Status While Waiting" section to master-health-monitoring.md
with:
- Example polling loop using sleep 30
- Reference to bin/ci-rerun-failures which already uses 30s intervals
- Rationale for why 30s is preferred over 180s

Updated avoiding-ci-failure-cycles.md to:
- Include example polling loop in the "After Pushing" section
- Emphasize the importance of 30-second vs 180-second polling

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

Author: justin808

.claude/docs/avoiding-ci-failure-cycles.md

@@ -201,8 +201,14 @@ Tested with:
 **After Pushing**:
 
 ```bash
-# Within 30 minutes: Check CI status
-gh pr view --json statusCheckRollup
+# Poll CI status every 30 seconds until completion
+while true; do
+  gh pr view --json statusCheckRollup --jq '.statusCheckRollup | group_by(.conclusion) | map({conclusion: .[0].conclusion, count: length})'
+  sleep 30
+done
+
+# Or use the automated tool (polls every 30s):
+bin/ci-rerun-failures
 
 # If failures occur:
 # 1. Check if they're new (not pre-existing)
@@ -211,6 +217,8 @@ gh pr view --json statusCheckRollup
 # 4. Don't push "hopeful" fixes
 ```
 
+**IMPORTANT: Poll every 30 seconds, NOT 180 seconds.** CI jobs typically complete in 3-15 minutes, so 30-second polling gives responsive feedback.
+
 **See**: [master-health-monitoring.md](master-health-monitoring.md)
 
 ## Red Flags: When to STOP and Test More

.claude/docs/master-health-monitoring.md

@@ -21,6 +21,31 @@ Within 30 minutes of your PR merging to master:
    - Don't assume "someone else will fix it"
    - You are responsible for ensuring your PR doesn't break master
 
+## Polling CI Status While Waiting
+
+**When actively monitoring CI, poll every 30 seconds (not 180 seconds):**
+
+```bash
+# Poll every 30 seconds while CI is running
+while true; do
+  gh pr view --json statusCheckRollup --jq '.statusCheckRollup | group_by(.conclusion) | map({conclusion: .[0].conclusion, count: length})'
+  sleep 30
+done
+```
+
+Or use the automated tool which polls at 30-second intervals:
+
+```bash
+bin/ci-rerun-failures  # Automatically waits for in-progress CI, polling every 30s
+```
+
+**Why 30 seconds?**
+
+- CI jobs typically complete in 3-15 minutes
+- 30-second polling gives responsive feedback without excessive API calls
+- 180 seconds (3 minutes) is too slow—you waste time waiting unnecessarily
+- GitHub API rate limits are 5000 requests/hour, so 30-second polls are fine
+
 ## When You Discover Master is Broken
 
 1. **Determine if it's from your PR:**