fix(ci): give falcon-hello a per-run port pair — fixed ports collide

[avrabe]

A red required check on a PR that could not have caused it.

Symptom

error: bind 127.0.0.1:14700: Address already in use (os error 98)
error: bind 127.0.0.1:14701: Address already in use (os error 98)

Hit on #342 — which changes one workflow file and cannot affect a UDP demo. main was green at the same commit range, so the failure was environmental.

Cause

falcon-hello binds a fixed UDP port pair (default 14700/14701). Several of these jobs can land on the same self-hosted runner concurrently. That is not a rare race: it happens whenever more than one PR is updated at once, which is routine here because branch protection is strict — every merge puts the remaining PRs BEHIND and they all get refreshed together. I triggered it by updating four PRs in one go.

Why fix rather than re-run

It presents as a genuine regression on an unrelated PR, so it costs a diagnosis cycle. And the thing that "works" is hitting re-run on a red gate — which is precisely the habit that gets a real failure clicked through.

Fix

scripts/falcon-hello-demo.sh already honours FALCON_HELLO_PORT_BASE (line 28). CI just never set it.

base = 20000 + (GITHUB_RUN_ID % 10000) * 2   ->  20000..39998
pair = base, base+1

Stepping by 2 is load-bearing — the demo uses base and base+1, so consecutive bases would still collide on one port.

GITHUB_RUN_ID is injected by Actions as a plain env var, so nothing is interpolated into the run: block.

Verified end-to-end, not by reading

$ FALCON_HELLO_PORT_BASE=20246 bash scripts/falcon-hello-demo.sh
[falcon-hello-demo] launching gcs on 127.0.0.1:20246
exit=0

Port range checked (20000–39999, valid), consecutive run ids confirmed 2 apart, YAML parses.

🤖 Generated with Claude Code

https://claude.ai/code/session_01HvusAXYbHLyv3uTzfBcMbG