Bump OpenTelemetry.Exporter packages to 1.15.3 (fixes GHSA-4625-4j76-fww9) and add E2E startup retry loop

[Copilot]

OpenTelemetry.Exporter.OpenTelemetryProtocol 1.8.1 carries a known moderate vulnerability (GHSA-4625-4j76-fww9 / CVE-2026-42191). Because the build runs with /warnaserror, the NuGet NU1902 audit warning was promoted to a build-breaking error.

Additionally, E2E tests on CI were flakily failing because the Console host occasionally crashes on startup with COMException (0x800700AA): The requested resource is in use. — a transient WebView2 resource conflict when host processes start in rapid succession.

Changes

• Directory.Packages.props

  • OpenTelemetry.Exporter.OpenTelemetryProtocol: 1.8.1 → 1.15.3 (patched version per advisory)
  • OpenTelemetry.Exporter.Console: 1.8.1 → 1.15.3 (brought in line with OpenTelemetry.Extensions.Hosting, already at 1.15.3)

• test/AdaptiveRemote.EndtoEndTests.TestServices/Host/AdaptiveRemoteHost.Builder.cs

  • Added a private HostStartupException class that wraps startup failures and carries the captured process stdout/stderr, enabling the retry logic to inspect what the crash reported.
  • StartWithSettings catch block now throws HostStartupException (wrapping the original) instead of rethrowing, so captured output is always available to callers.
  • Start() now retries up to 3 times when the host crashes with "The requested resource is in use." in its output. Each retry logs a warning and waits a random 1–3 s before the next attempt. All other failure types propagate immediately.

[github-actions]

Test Results

540 tests  +540   540 ✅ +540   5m 30s ⏱️ + 5m 30s
  8 suites +  8     0 💤 ±  0 
  8 files   +  8     0 ❌ ±  0 

Results for commit d5468d5. ± Comparison against base commit 2476d59.

♻️ This comment has been updated with latest results.

[jodavis]

@copilot the E2E tests failed, but it's a known issue. Sometimes on CI, the application fails to start because it's being shut down and restarted frequently. This is not a product bug, so I want you to add a retry loop to the test host start up code. If the app host crashes right away on startup, with this specific exception message, just try starting it again. It's more important to run the tests and catch real bugs than it is to straighten out this flaky edge case.

[Copilot]

@copilot the E2E tests failed, but it's a known issue. Sometimes on CI, the application fails to start because it's being shut down and restarted frequently. This is not a product bug, so I want you ...

Done in d5468d5. AdaptiveRemoteHost.Builder.Start() now retries up to 3 times when the host crashes on startup with "The requested resource is in use." in its captured output (the COMException 0x800700AA from WebView2). Each retry logs a warning and waits a random 1–3 s before trying again. The retry only triggers for that specific crash — other failures (Inconclusive, process-launch errors, timeout, etc.) propagate immediately as before.