Tailscale routes missing in ephemeral runners

Hello team!

We are using the action to run a simple curl request in our tailnet, and it has been failing randomly. After some digging, we noticed the following:

• Tailscale ping does work (for the DN and IP requested). We tried manually, and through the the action ping: <dn>,<ip>
• Once a runner is failing, it does not recover by itself (even after 10 minutes). Curl still fails after that.
• We noticed some routes are missing ip route show table all |grep tailscale only shows the following:

fe80::/64 dev tailscale0 proto kernel metric 256 pref medium
local fe80::ba:54d9:b084:82e9 dev tailscale0 table local proto kernel metric 0 pref medium
multicast ff00::/8 dev tailscale0 table local proto kernel metric 256 pref medium

While on a working runner, we do have the desired route showing.

Hotfix: a simple tailscale down && tailscale up in the action does solve the issue.

Current action is as follow:

      - name: Setup Tailscale
        uses: tailscale/github-action@53acf823325fe9ca47f4cdaa951f90b4b0de5bb9 # v4.1.1
        with:
          version: TAILSCALE_VERSION
          oauth-client-id: <some-client-id>
          oauth-secret: <some-client-secret>
          tags: tag:<some-tag>
          ping: <some-internal-ips>,<some-internal-dn>

Thank you!
Jules

Start by reproducing the workflow with the shown Tailscale setup and compare ip route show table all | grep tailscale on failing and working runners. Check the action's setup flow around route initialization, then verify that the expected route is present and curl succeeds without running tailscale down && tailscale up.