What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
You can test a workflow that sleeps for three days without letting three real days pass: freeze the test clock, keep background queue work from running on its own, and advance the workflow deliberately. The important part is to test what happens before and after its wake time—not just whether a step reports a three-day delay.
Steve McDougall’s August 2026 article, “Seven Days in Ten Milliseconds”, uses a Laravel abandoned-cart reminder to show how controlled time can make long waits testable. The title is a description of test-time behavior, not a production performance benchmark: it does not establish that a real workflow completes seven days of work in ten milliseconds.
As an Amazon Associate I earn from qualifying purchases.
What a sleeping workflow should do
In the example, a workflow sends an initial cart reminder, waits three days, sends a final reminder, and then ends. The delay is 259,200 seconds. A test that checks only that duration can pass even if the workflow never wakes correctly or its later action fails.
The model described in the article treats sleep as persisted workflow state: the instance is marked sleeping and has a future wake time. A delayed job is expected to move it forward when that time arrives. This lets a test examine both the parked state and the behavior after waking instead of waiting in real time.
#1 Best Overall
Drive the workflow with a controlled clock
- Freeze test time. Establish a known starting point so the workflow’s wake time can be evaluated predictably.
- Fake the queue. Prevent background work from advancing the workflow unexpectedly; the test should decide when to call the engine’s advance operation.
- Start the workflow and inspect its first step. Confirm that the initial reminder is sent and the instance parks in a sleeping state with the expected future wake time.
- Advance the clock by two days, then try to advance the instance. It should remain asleep, and the final reminder should not be queued. This negative check catches an early transition that a delay-value assertion would miss.
- Move test time beyond the three-day wake time and advance again. Confirm that the final reminder is sent after the wait.
- Advance once more and verify completion. In the engine behavior described by McDougall, one call processes one step. The test therefore needs a separate call to move past the final step and complete the instance.
Assert on behavior, not only the delay
A duration assertion can confirm that a step requested a three-day sleep, but it does not prove that the scheduler will resume the workflow, that the delayed job will be handled, that workflow context survives the wait, or that the later side effect occurs at the right time. The stronger test checks the sequence: the instance sleeps, an early advance leaves it asleep with no later reminder, an advance after the wake time sends that reminder, and the instance eventually completes.
- Before the wake time: assert the workflow has not advanced and the later reminder is absent.
- After the wake time: assert the expected reminder side effect occurs.
- At the end: assert the instance reaches completion, rather than stopping at the final side effect.
What the “ten milliseconds” framing does—and does not—mean
The example’s three-day delay is simulated by moving the test clock, so the test does not need to wait three calendar days. The article’s headline is not a measured result across machines, test suites, or production environments, and it should not be read as a guarantee that any workflow engine will execute a seven-day process in ten milliseconds. The practical point is narrower: when a test controls time and advancement, it can exercise long-delay logic without incurring the modeled wait.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Recommended Free Tools
