[HBartosch]

🚀 Pull Request: Routya.Core

---

📋 Summary

The CI job summary and the README both publish allocation figures under near identical labels, but
they measure different things. The first post merge run reported
Send with a Singleton handler: 104 B while the README says
Routya Send, Singleton handler: 808 B. Both are correct, but nothing explained the eightfold
difference, so a reader comparing them would reasonably conclude one was wrong.

Documentation and test labels only. No behaviour, no published figure and no budget changes.

🔗 Related Issues (If Any)

Closes #___
Linked to: #___

🛠️ Changes Made

• Request/notification handler logic
• Pipeline behavior changes
• Dispatcher performance or safety
• DI registration improvements
• Unit tests added or updated
• Benchmarks updated (if performance-related)

Only the test scenario label strings changed. No test logic, no assertions and no budgets were
touched, which the diff shows directly: every changed line in both test files is an
AssertWithinBudget(...) description argument.

✅ Checklist

• Code compiles and passes unit tests
• No breaking changes introduced (or documented)
• Docs/readme updated if needed
• Follows Routya code style and conventions
• Scoped vs root resolution behavior remains correct

The last item is ticked on the basis that no dispatch code is touched by this branch and the full
suite passes unchanged, not because new scope testing was added.

🧪 Manual Testing Notes

Rendered the CI job summary locally by setting GITHUB_STEP_SUMMARY and running both suites, to
confirm the new labels read correctly in the table that appears on the workflow run page:

| Send, Singleton handler, dispatch only                      | 104 B | 128 B  |  24 B |
| PublishAsync, 2 Singleton handlers, dispatch only           |  32 B |  64 B  |  32 B |
| SendAsync, Scoped handler, includes dispatch scope          | 544 B | 640 B  |  96 B |
| PublishParallelAsync, 2 Scoped handlers, scope per handler  | 824 B | 1024 B | 200 B |

Verified the README's new claim against the benchmark source rather than assuming it:
Routya.Benchmark registers exactly two Routya pipeline behaviours and two equivalent MediatR
behaviours per provider, and each benchmark method calls CreateScope() and
GetRequiredService<IRoutya>() inside the measured region. So "the full cost of serving one
request" and "both libraries are measured the same way" are both accurate as written.

Automated, all run and passing:

Suite	Tests
Routya.Test	64
Routya.SourceGenerators.Test	31
Routya.SourceGen.Test	23

💬 Additional Context

Why the two sets of figures differ

	README table	Allocation budget tests
Measures	The whole cost of serving one request	Dispatch on its own
Includes	CreateScope(), GetRequiredService, two pipeline behaviours, the handler	Dispatcher already resolved, no behaviours
Purpose	Comparing libraries, so MediatR is measured identically	Catching a regression, so Routya is isolated from the container
Singleton Send	808 B	104 B

Both units are the right one for their job. The problem was only that neither said so.

What changed

• Budget test scenarios are now self describing, for example Send, Singleton handler, dispatch only
  and PublishParallelAsync, 2 Scoped handlers, scope per handler, so the CI summary explains itself
  without reference to anything else.
• The README states what its figures include, and states the relationship outright: a Singleton
  Send is 104 B of dispatch inside 808 B of total work.

Strengthened claim, now evidenced

The README previously asserted that allocations "hold on your hardware as well as ours". The first
post merge CI run returned all nine figures identical to the byte against measurements taken on
a different machine and operating system, so the README now says that was verified independently
rather than asserting it. That is a materially stronger claim than the timings it replaced could
make: the MediatR baseline, running unchanged code, has been observed drifting by a third between
two runs on a single machine.