fix: isolate tooling telemetry and test pregenerated Wolverine handlers

[PANiXiDA]

The Host initialized OpenTelemetry during functional tests and Wolverine code generation, while tests could silently compile handlers dynamically instead of checking the generated code used by deployment. Skip observability in the Test environment and for the codegen command, and add a strict pregenerated-handler test path.

Identity and Compendium web factories explicitly select the Host assembly. WOLVERINE_PREGENERATED=1 selects Static loading; ordinary tests retain Auto. A dedicated CI workflow generates handlers, recompiles them and runs the complete solution with Static. Production loading settings are unchanged.

The manually dispatched benchmark workflow compares clean-output build/test cycles on 20 separate GitHub runners. Each runner warms both variants and measures one pair; order is balanced 10/10. Total time includes the bootstrap Host build, codegen write, recompilation and all test projects with their containers. Restore, image downloads and warmups are excluded equally.

Mean elapsed time	Auto without generated types	Codegen + Static	Change	Paired 95% CI
Complete build/codegen/test cycle	181.227 s	183.145 s	+1.06%	-0.33% to +2.56%
Test execution	155.632 s	144.718 s	-7.01%	-8.17% to -5.78%

Test execution saved 10.914 seconds, while additional preparation cost 12.832 seconds. The complete cycle shows no statistically significant improvement. Retain the static path to validate pregenerated handlers, rather than claiming an end-to-end speed optimization.

Validation:

• Benchmark run 37641341275 passed: 40 measured and 40 warmup runs, each with all 1210 tests passing, with no measured samples excluded.
• Local full Release Static preflight passed 1210/1210 tests.
• Negative control without generated files failed with MissingPreBuiltTypesException for 45 handlers, confirming no dynamic fallback.
• Functional tests verify that Test hosts omit OpenTelemetry providers and Development hosts still register them.
• Targeted dotnet format validation, actionlint and git diff checks passed.
• Final commit 5b143b5 passed the complete CI and Sonar Quality Gate and the dedicated pregenerated-handler test workflow. Subsequent script style/comment corrections preserved the recorded statistics exactly.

Method, full results and raw measurements are committed with the benchmark scripts. This experiment holds the observability flag constant; it measures pregeneration plus Static, including its generated handler registry. It does not measure deployed containers or the latency of the existing parallel CI matrix.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-07T14:20:16.352052Z	a7daa45	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.