flake: e2e externalAuth.spec.ts

[johnstcn]

Problem

See #212 for context.

==> Errors
/home/runner/work/coder/coder/site/e2e/tests/externalAuth.spec.ts:15:
  13 | import { beforeCoderTest, resetExternalAuthKey } from "../hooks";
  14 |
> 15 | test.beforeAll(async ({ baseURL }) => {
     |      ^
  16 | 	const srv = await createServer(gitAuth.webPort);
  17 |
  18 | 	// The GitHub validate endpoint returns the currently authenticated user!
"beforeAll" hook timeout of 30000ms exceeded.

It looks like the hardcoded port (50516) can be unavailable for more than 30s.

Proposed Solution

1. Simply bump the timeout (chore(site/e2e): increase timeout for waitForPort from 30s to 60s coder#16528)

2. Refactor the e2e tests to dynamically create a port.

[ethanndickson]

Spotted again https://github.com/coder/coder/actions/runs/14104410647/job/39507590233

[aslilac]

disabling test for now until someone has the bandwidth to patch

[aslilac]

• I don't think bumping the timeout will reduce the flakes, it'll just make runs take longer
• It's hard to just pick random ports for this because the ports have to be configured in advance as env variables
• I'm also not really convinced that picking a random port would actually solve the issue, because it's already a random port number that we only reference in exactly one place. nothing else in our ci environment should be randomly grabbing it.

it seems like no matter what the port number is, sometimes playwright runs our code twice, and we can't listen to the same port a second time.