ci(e2e): run e2e specs with 3 workers on CI

[gabrielseco]

Summary

E2E specs run 3 at a time on CI instead of one after another. The e2e job drops from about 2.5 minutes to about 1.5.

Stacked on #1485. Review that first; this PR's net diff is only the worker count.

Why

workers: process.env.CI ? 1 : undefined is the line Playwright's scaffold generates, and it has been there since the original "move to playwright" commit. It doesn't look like a deliberate choice. The setup already supports parallel runs:

• fullyParallel: true is set.
• CI runs the browsers against a deployed Vercel preview, so the runner isn't also serving the app.
• Each spec creates its own employment/contractor with Date.now() names and emails, so specs don't obviously share data.

This PR first tried the official Playwright Docker image as well, to skip the apt install. It didn't help. Across 23 recent runs the apt step took 11–33s (the 2m41s that prompted it was a one-off), and starting the container costs about 28s. On top of that, someone would have to bump the image tag by hand. It's reverted here, so only the worker change remains.

What changed

Toggle details

• example/playwright.config.ts: CI workers 1 → 3 (ubuntu-latest has 4 vCPUs). Removed the scaffold's "Opt out of parallel tests on CI" comment, which would now be wrong.
• Local runs are unchanged (undefined = Playwright's default).
• Branch history: container commit → workers commit (ci(e2e): run e2e specs with 3 workers on CI #1487) → revert of the container commit. Squash merge makes this a single config change.
• Main risk: sandbox load. 3 specs now seed data against the sandbox at the same time. If seeding gets slow or rate-limited, it would show up as retry # lines in the list output, not as clean failures.
• No SDK or public API changes.

Screenshots

N/A

Related Resources

• Stacked on ci(e2e): only surface failures in Playwright CI output #1485
• ci(e2e): run e2e specs with 3 workers on CI #1487 (workers commit, merged into this branch)

Testing

Measured E2E job timings:

Run	Job total	Tests
#1485, 1 worker	2m35s	1.7m
container + 1 worker	2m34s	1.6m
container + 3 workers	1m36s	40s, no retries

• 3 workers: all 9 specs pass, no retries or flaky tests
• Run on this branch (no container): all 9 pass in 39s, no retries
• Feature flag: N/A

🤖 Generated with Claude Code

---

Note

Low Risk
Only affects CI test execution parallelism in the example app; no production or SDK behavior changes.

Overview
CI E2E parallelism: On CI, Playwright now runs with 3 workers instead of 1, so test files can run in parallel again rather than being forced serial on the runner.

The old comment about opting out of parallel tests on CI was removed to match the new setting. Local runs are unchanged (workers stays undefined outside CI).

^{Reviewed by Cursor Bugbot for commit 0949451. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

Deploy preview for remote-flows ready!

Project:	remote-flows
Status:	✅  Deploy successful!
Preview URL:	https://remote-flows-aigg7fyt0-remotecom.vercel.app
Latest Commit:	0949451

Deployed with vercel-action

[github-actions]

Deploy preview for adp-cost-calculator ready!

Project:	adp-cost-calculator
Status:	✅  Deploy successful!
Preview URL:	https://adp-cost-calculator-j61alh9zi-remotecom.vercel.app
Latest Commit:	0949451

Deployed with vercel-action

[github-actions]

📊 Coverage Report

⚪ Coverage unchanged

Metric	Current	Previous	Change	Status
Lines	89.31%	89.31%	0%	⚪
Statements	88.86%	88.86%	0%	⚪
Functions	88.07%	88.07%	0%	⚪
Branches	80.37%	80.37%	0%	⚪

Detailed Breakdown

Lines Coverage

• Covered: 5249 / 5877
• Coverage: 89.31%
• Change: 0% (0 lines)

Statements Coverage

• Covered: 5347 / 6017
• Coverage: 88.86%
• Change: 0% (0 statements)

Functions Coverage

• Covered: 1402 / 1592
• Coverage: 88.07%
• Change: 0% (0 functions)

Branches Coverage

• Covered: 3250 / 4044
• Coverage: 80.37%
• Change: 0% (0 branches)

✅ Coverage check passed