Add diagnostic logging to MCP gateway health check failures

[Copilot]

MCP gateway failures at workflow step 24 lacked diagnostic information to identify which server failed and why. The check_mcp_servers.sh script would report "server(s) failed" without details on failure type, retry attempts, or specific servers.

Changes

Enhanced check_mcp_servers.sh with structured diagnostic output:

• Retry visibility: Logs each attempt with timeout progression (10s → 20s → 30s)
• Error classification: Distinguishes connection refused, timeout, DNS failure, HTTP codes, JSON-RPC errors
• Failure context: Per-server output includes URL, last error, retry count
• Actionable summary: Failure counts, common causes, log file paths

Example Output

Before:

✗ serena: failed to connect
ERROR: 1 server(s) failed

After:

✗ serena: failed to connect
  URL: http://host.docker.internal:80/servers/serena
  Last error: Connection timeout
  Retries attempted: 3

ERROR: 1 of 6 server(s) failed connectivity check
Succeeded: 5, Failed: 1, Skipped: 0

Common causes:
  - MCP server container failed to start or crashed
  - Network connectivity issues between gateway and server
  - Server initialization taking longer than expected (>30s)

Check logs: /tmp/gh-aw/mcp-logs/stderr.log

This enables rapid differentiation between transient timing issues and persistent connectivity failures.

Original prompt

Reference: https://github.com/githubnext/gh-aw/actions/runs/21314137115/job/61354509874#step:24:1

Debug MCP gateway failure

Custom agent used: agentic-workflows
GitHub Agentic Workflows (gh-aw) - Create, debug, and upgrade AI-powered workflows with intelligent prompt routing

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.