Scrub the last fleet identifier from main

[erikdarlingdata]

main is the repo's default branch and is swept separately from dev. The two have genuinely divergent history, so the identifier scrubs merged to dev ([#2900], [#2903]) never reached it — four locations on main still named a real tenant server.

What changed

file	was	now
Darling.Tests/DarlingPeerDisclosureTests.cs:385	display_name fixture	omega
Darling.Tests/DarlingPeerDisclosureTests.cs:413	matching assert	omega
Darling.Tests/StatementSplitTimingTests.cs:79	short name in comment	omega-01
PerformanceMonitor.Collectors/CollectorContext.cs:253	short name in comment	omega-01

omega is not a fresh choice — it is the slug [#2900] reconciled this server onto, it is already in FleetIdentifierScrubTests.SyntheticSlugs, and these four lines now match origin/dev byte-for-byte (verified by diffing slug usage per file). Every measurement in the touched comments is unchanged.

Line endings preserved per file (CRLF counts asserted before/after).

Same precedent as 26a4e14, "Scrub fleet identifiers from main, which is what the repo shows by default".

Why a PR and not a direct commit

26a4e14 went straight onto main. This is the same change shape, but a PR leaves the diff reviewable and revertable, and there is no rush now that the issue/PR-body exposure is closed.

🤖 Generated with Claude Code

[claude]

Reviewed. This is a pure identifier scrub — the old name/the old short form → omega/omega-01 in test fixture data and doc comments only, no logic changes.

• Confirmed omega is already registered in FleetIdentifierScrubTests.SyntheticSlugs, so the guard test will still pass and won't flag these as new leaks.
• Confirmed no other the old name occurrences remain in the tree (grep across .cs/.sql finds none outside unrelated substring matches like "display"/"DisplayRoot").
• CollectorContext.cs lives in the shared PerformanceMonitor.Collectors project, not a Lite- or Darling-specific tree, so there's no parity counterpart to update.
• No T-SQL touched, so the CONTRIBUTING.md style rules (aliasing, OPTION (RECOMPILE), etc.) don't apply here.

No correctness, security, or performance concerns. LGTM.

[erikdarlingdata]

Closing. This can't land as shaped — check-branches fails by design, because the workflow only permits PRs to main from dev, and this targets main from a topic branch. That's the policy working, not a flaky check.

It also isn't needed. The equivalent changes are already on dev (#2900, #2903), those commits are simply not in main yet, and they arrive with the normal release merge — dev already holds the corrected blobs for every file this PR touches. No separate action required.

The verification work here was sound and the slug choice was right (matching what #2900 reconciled to, rather than picking fresh). Nothing wrong with the change itself — just the branch it targeted.