Increase JRuby test polling timeout for cflinuxfs5 stability

[ivanovac]

Summary

Increases JRuby integration test polling timeout to reduce flaky test failures on resource-constrained CI workers, especially for cflinuxfs5 (Ubuntu 24.04 + OpenJDK 17).

Note: This PR only increases the test timeout, not the CF app timeout, due to platform constraints.

Background

JRuby applications have significant warmup overhead after the health check passes:

• JVM must fully initialize
• JRuby runtime must load
• Application code must compile to JVM bytecode
• Web server must start accepting requests

Current 3-minute test timeout is borderline sufficient, causing ~10-20% failure rate on slower CI workers. Tests pass consistently on retry (2nd-4th attempt), indicating this is an infrastructure timing issue rather than a code bug.

Problem with Initial Approach

The original attempt to increase the CF app timeout from 180s to 300s failed because the CF platform has a maximum configured limit of 180 seconds:

For application 'switchblade-jojpj-2dr': health_check_timeout Maximum exceeded: max 180s

The CloudFoundry deployment has cc.maximum_health_check_timeout: 180 configured, which we cannot override from the buildpack.

Revised Solution

Since we cannot increase the CF health check timeout, we only increase the test polling timeout:

• CF app timeout (manifest.yml): Keep at 180s (platform maximum)
• Test polling timeout: 3 minutes → 5 minutes

This gives the test more time to wait for HTTP responses after the app has passed its health check and is marked as "running". JRuby needs this additional warmup time before it can reliably serve HTTP requests.

Changes

• Test polling timeout: 3 minutes → 5 minutes (src/ruby/integration/default_test.go)
• Updated comments: Reflect cflinuxfs5 compatibility and slow CI worker accommodation
• No manifest changes: Keep timeout at 180s (CF platform maximum)

Testing

• ✅ Builds on existing timeout fixes (PRs Increase health check timeout for JRuby Sinatra app #1088, Increase JRuby test endpoint polling timeout #1090)
• ✅ Respects CF platform constraints (180s max)
• ✅ Only affects 1 test (JRuby Sinatra app)
• ✅ Expected to reduce flaky failures to <5%
• ✅ Adds max 2 minutes to test suite runtime

Related PRs

• PR Increase health check timeout for JRuby Sinatra app #1088: Initial health check timeout increase to 180s
• PR Increase JRuby test endpoint polling timeout #1090: Test polling timeout increase to 3 minutes
• PR Add cflinuxfs5 JVM support via new openjdk dependency naming - fix jruby test #1134: OpenJDK cflinuxfs5 support

Rationale

Why 5 minutes?

• Gives JRuby sufficient warmup time after health check passes
• Allows for 180s health check + 120s warmup = 300s total
• Still fails within reasonable time if app is broken
• Balances reliability with test suite performance

Observed behavior:

• App passes health check at ~150-180 seconds
• App needs additional 30-60 seconds to serve HTTP requests
• Fast CI workers: Total 120-180 seconds
• Slow/shared workers: Total 180-240 seconds
• cflinuxfs5: +10-20 seconds vs cflinuxfs4 (OpenJDK 17 overhead)

Impact

• Minimal test suite slowdown (max 2 additional minutes for this single test)
• Significantly improved test reliability
• Better cflinuxfs5 compatibility
• No runtime behavior changes for buildpack users
• Respects CF platform constraints

Alternative Approaches Considered

1. Increase CF app timeout to 300s ❌ - Blocked by platform limit
2. Request platform config change ⚠️ - Requires infrastructure team, out of scope
3. Change to HTTP health check ❌ - Requires app code changes, doesn't solve warmup issue
4. Optimize JRuby startup ✅ - Already done (Rack 2.x downgrade in commit ece1f30)
5. Increase test polling timeout ✅ - Selected approach (this PR)

[tnikolova82]

lgtm