fix: timing-based test assertions are inherently flaky

[thomhurst]

Description

Several tests verify parallel/non-parallel execution by checking DateTime overlaps, which are inherently fragile:

ParallelTests.cs and NotInParallelTests.cs

Use DateTimeRange overlaps to verify parallel/sequential execution. DateTime precision depends on system clock granularity. On slow CI systems, tests might complete too quickly for overlap detection.

TimeoutTests1.cs (Lines 18-19)

Tight timing assertion expecting test to complete between 4-60 seconds. Fragile on resource-constrained CI.

AdaptiveParallelismTests.cs

CPU-intensive loop using DateTime.UtcNow comparisons (system clock dependent). 1000 repetitions × 1s delay = 16+ minutes per test method.

Suggested Fix

• Replace DateTime overlap checks with Interlocked concurrent execution counters
• Use event-based synchronization (TaskCompletionSource) instead of hard timeouts
• Widen timing windows with documented rationale
• Separate AdaptiveParallelismTests into a [Category("Performance")] suite

[github-actions]

This issue is stale because it has been open 30 days with no activity. Remove stale label or comment or this will be closed in 5 days.