fix(ssh): keep github.com connection alive through slow pre-push hooks

[catomean]

Summary

• git push opens the SSH connection to github.com before running the pre-push hook, then leaves it idle for the hook's entire duration.
• Under heavy concurrent machine load (~80 sessions sharing this box), app-repo pre-push hooks (typecheck + full test suite) routinely took 5-10+ minutes — long enough that GitHub's idle-connection timeout closed the SSH session mid-hook.
• Symptom: the hook suite prints "🎉 All pre-push checks passed! Safe to push." and then... nothing. The branch never lands on the remote, with only a stray "Connection to github.com closed by remote host." buried mid-log to explain it. Reproduced 4x in one session on orangecat before diagnosing via GIT_TRACE=1.
• Fix: ServerAliveInterval 30 / ServerAliveCountMax 10 on the github.com SSH host keeps the connection alive with active keepalives during that idle window instead of relying on it staying up passively.

Test plan

• Diagnosed root cause via GIT_TRACE=1 GIT_SSH_COMMAND="ssh -vvv" git push — confirmed the SSH session opens before the hook and drops during it
• bash -n .ssh/config (repo's own CI check)
• Watch a subsequent orangecat/fleetcrown push under load land cleanly on the first try

🤖 Generated with Claude Code

https://claude.ai/code/session_01Y9rKLxddothnXEtY6KDziN