Up to standards ✅

_{NEW Get contextual insights on your PRs based on Codacy's metrics, along with PR and Jira context, without leaving GitHub. Enable AI reviewer
TIP This summary will be updated as you push new changes.}

Code Review — PR #6316: Update Mock Benchmark Results

This is an automated daily PR that refreshes mock benchmark documentation by updating timestamps (2026-06-25 → 2026-06-26), regenerating benchmark numbers in .md files, and syncing the source .json data files. No application logic is touched; all 15 changed files are in docs/.

Finding 1 — Schema inconsistency: "Median" field appears only in Setup category

Files: docs/static/benchmarks/mocks/Setup.json, docs/static/benchmarks/mocks/latest.json (Setup section)

The new Setup.json adds a "Median" field to every result entry (e.g., "Median": "504.8 ns"), but no other category JSON (Callback.json, CombinedWorkflow.json, Invocation.json, MockCreation.json, Verification.json) includes this field. The latest.json faithfully mirrors this inconsistency — Setup entries have Median, everything else doesn't.

Why this matters: If any tooling (the Docusaurus site, a CI comparison script, or a future consumer) iterates over benchmark results and expects a uniform schema per category, it will either silently ignore the median for non-Setup categories or crash when it's missing. This also makes the published docs misleading — readers can't compare median performance across categories.

Suggested fix: Either emit Median for all categories when it's available in BenchmarkDotNet output, or strip it from Setup so the schema stays uniform. A schema validation step in the workflow that generates this PR would catch this automatically on future runs.

Finding 2 — "Gen2" field removed inconsistently across files

Files: docs/static/benchmarks/mocks/Callback.json (and latest.json Callback section) vs. MockCreation.json (which retains "Gen2": "-")

In the Callback category, this PR removes "Gen2": "-" from all result entries. MockCreation still carries the field. The result is that two JSON files with the same stated schema now differ structurally — some objects have two GC-generation fields, others have three.

Why this matters: Same risk as above: tooling or display logic that iterates over GC fields will behave differently per category without any indication this is intentional. If Gen2 collections are always - for Callback, the cleaner fix is to emit it consistently (or omit it consistently) rather than letting BenchmarkDotNet's run-to-run output variation silently produce different schemas.

Observation — Dramatic CombinedWorkflow performance swings

The CombinedWorkflow benchmark shows very large day-over-day deltas for the proxy-based libraries:

• Moq: 400.863 μs → 124.317 μs (−69%)
• NSubstitute: 17.322 μs → 8.903 μs (−49%)
• FakeItEasy: 18.466 μs → 9.180 μs (−50%)

These improvements are far outside normal benchmark noise (< 5%) and suggest the prior day's run had significant environmental interference (JIT warm-up, shared CI runner load, etc.) rather than genuine performance changes. The resulting Mermaid chart y-axis shifted from 0 --> 482 to 0 --> 150, which would visually mislead a reader comparing two days' charts.

This isn't a bug in the PR itself — it faithfully reproduces the CI output — but it's a signal that the benchmark automation workflow would benefit from a stability gate (e.g., reject or flag results where a library's mean changes by > 20% vs. the rolling 7-day average) before auto-committing. Without that, the daily-updated docs can quietly reflect noisy runs as authoritative results.

Overall the automated pipeline is working correctly and the files are internally consistent within each run. The schema divergence between categories (Median/Gen2) is the most actionable item; the rest is a suggestion for improving benchmark reliability.