Stabilize recurring CI flake cluster (#4310)

[jeremydmiller]

Closes #4310.

Root causes

Three independent issues feed the same flake pattern. Each is localized; this PR addresses all three.

1. Shared static random sequence in Target.cs

private static readonly Random _random = new Random(67) is consumed by ~80 tests across LinqTests + DocumentDbTests. Every Target.Random() / GenerateRandomData() call advances the shared sequence, so a test's effective data depends on which sibling tests consumed before it. xUnit discovery order isn't perfectly stable across CI workers and TFMs — a small order shift consumes a different slice and silently flips assertions that depend on exact counts or distributions.

Fix. Introduce Target.ResetRandomSeed(int) and call it in the tests that genuinely depend on specific data:

• Bug_605_unary_expressions_in_where_clause_of_compiled_query (3 facts)
• Bug_3337_select_page.try_it_out
• query_against_child_collections.buildUpTargetData (covers can_query_on_enum_properties and other variants)

Also tightened Bug_605 assertions: .ShouldBe(15) was a fragile hardcode; the real point of the test is "compiled query == inline LINQ query". Compare against expected.Count.

2. DateTimeOffset.UtcNow inside a shared LINQ expression

child_collection_queries.cs:67 registered:

@where(x => x.Children.Any(c => c.NullableDateOffset <= DateTimeOffset.UtcNow));

That expression runs in BOTH the in-memory LINQ-to-objects "expected" provider AND the LINQ-to-SQL "actual" provider. Each evaluates DateTimeOffset.UtcNow at its own moment. Target.NullableDateOffset values are ±60 seconds of "now" from random data; values within microseconds of either provider's "now" can land on opposite sides of <= and disagree.

Fix. Capture asOf = DateTimeOffset.UtcNow.AddDays(1) once in the static ctor and use that as the boundary. The expression now embeds a constant timestamp both providers see identically. AddDays(1) is well beyond the test data range so the predicate remains meaningful.

3. Ordering assumptions on server-generated Guids

Bug_4282 did ids.ShouldHaveTheSameElementsAs(doc1.Id, doc3.Id) after OrderBy(x => x.Id). Server-generated Guids don't necessarily sort in declaration order. Switched to set-membership: Count == 2 + ShouldContain per expected id.

4. Defensive: ShouldBeEqualWithDbPrecision tolerance

The helper used to round both sides to 100µs with truncation, then ShouldBe. The math works in the common case but the assertion was fragile on loaded-runner clock-comparison edges. Switched to a 1ms tolerance check (well above PostgreSQL's worst-case 9-tick truncation, still tight enough to catch real differences), with a clearer failure message.

Test plan

• 5x consecutive stress runs of LinqTests.Bugs (178 tests each) — 0 failures
• 123 tests across Bug_605 + Bug_4282 + Bug_3337 + query_against_child_collections + child_collection_queries — all pass
• Bug_2283 in DocumentDbTests passes

🤖 Generated with Claude Code