•

Recurring Work and Daylight Saving: A Clock Contract

An unmarked metronome and metal spool linked by a ribbon crossing an arc and a doubled branch, representing deliberate recurring-work clock rules.
What’s in this article?

    Decide what recurring work should do when a local time disappears or happens twice, then prove the scheduler and receiving team follow the same rule.

    A daily review is due at the same local time. After a clock change, one team sees no task while another sees two start messages. Both assume the scheduler has decided what the business intended. That is the wrong place to leave the decision.

    A recurring-work clock contract defines which occurrences should exist, how their scheduled times are interpreted and who accepts the resulting work. It connects an operating rule to tested system behavior, rather than treating a successful trigger as proof of a completed obligation.

    This blueprint is for routine internal reviews and service coordination. It does not authorize skipping a safety check, contractual obligation or regulated activity. Those cadence and exception decisions belong to the responsible policy owner.

    Choose the business clock before configuring the timer

    “Daily” can mean different things. Choose the meaning that serves the work, then identify whether the proposed scheduler can deliver it.

    • Local wall-clock cadence: start at a stated time in a named business zone, such as 08:00 America/Chicago. The local time matters more than a constant elapsed interval.
    • Elapsed-time cadence: start at a defined interval from an agreed anchor, such as every 24 hours. Do not assume this will preserve the same local hour across an offset change.
    • Completion-relative cadence: make the next occurrence due after accepted completion. Specify whether the anchor is submission, approval or another event; they are not interchangeable.

    The distinction is visible in Amazon EventBridge Scheduler’s schedule documentation: a rate using days represents a 24-hour duration, while timezone-based cron schedules have specific daylight-saving behavior. A cron occurrence at a nonexistent local time is skipped; a repeated local time produces one invocation. Those are that service’s rules, not universal business permissions.

    UTC is useful for recording an unambiguous instant. It does not answer whether the business needs “08:00 at the site” or “24 hours since the prior anchor.” Store the business rule as well as its resolved timestamp.

    Write one clock contract for each recurring series

    The work object at this stage is a series-release case: the proposal, approvals and tests for one recurring-work rule. Individual occurrences remain separate records. A change to the series must not silently rename older work or erase its original due time.

    Minimum recurring-work clock contract
    DecisionWhat the owner recordsAcceptance evidence
    Work and clockSeries ID, purpose, covered period, cadence type and anchorReader can distinguish the reporting period from the start time
    Governing zoneNamed business timezone and authoritative site/team; not the viewer’s device settingLocal label, offset and resolved UTC instant agree for test cases
    Missing local timeAuthorized choice: omit, move to an approved valid time, or hold for reviewExpected occurrence and disposition are explicit
    Repeated local timeOne occurrence at a selected instant, or two genuinely separate obligationsEach intended occurrence has one identity and resolved instant
    Execution boundariesPermitted start window, late-arrival rule, overlap rule and outage treatmentLate, retried or still-running work cannot silently create another obligation
    Authority and releaseBusiness approver, operator, receiving owner, rule version and effective boundaryApproved rule, observed configuration and accepted handoff match

    Do not infer an authorized omission from a scheduler’s skipped invocation. If the business requires work on every relevant date, choose and test an alternative valid start or an owned exception route before activation.

    Keep a rule version for traceability, but do not automatically give an already-existing obligation a new identity merely because its rule changes. Map the old and new representations at the effective boundary so both cannot dispatch the same work.

    Separate schedule identity, occurrence identity and attempts

    A schedule definition says when a series may produce work. An occurrence identifies the particular business obligation. An execution attempt records a try at delivering or performing it. These need different identities.

    For a once-per-business-date review, a series ID plus the intended business date and relevant site may identify the obligation. For a design that intentionally requires both instances of a repeated local time, include an unambiguous distinction such as the resolved instant. “01:30” alone cannot distinguish them.

    Retries retain the occurrence identity. A new attempt ID is not permission to create another task, send another consequential instruction or count another completion. Before retrying an uncertain dispatch, reconcile the original occurrence and receiving record through the system’s supported mechanism.

    Schedule uniqueness and occurrence-level duplicate protection are separate requirements. Even one schedule can produce retried delivery attempts. Conversely, suppressing duplicate attempts cannot repair two active schedule definitions that represent the same series.

    Test the missing hour and repeated hour as different cases

    Write the expected result before running a controlled test. The following matrix is a recommended design asset, not a claim that every platform supports each choice.

    Clock-transition acceptance matrix
    CaseDecision to testEvidence required
    Ordinary date before and after the changePreserve the chosen local time or elapsed intervalExpected local/UTC pair, actual trigger and accepted task
    Gap: configured local time does not existApply the approved omit, valid-time or review policyExplicit disposition; no invented completion or silent coverage loss
    Fold: local time occurs twiceSelect one resolved instant, or authorize two distinct obligationsCorrect occurrence count, offsets and receiving records
    Retry or late deliveryReconcile the original occurrence before another actionLinked attempts and one intended business effect per obligation
    Previous work still openApply approved hold, parallel-work or other supported overlap ruleOwner, dependency and next permitted action
    Rule changes at the release boundaryAssign each affected occurrence to exactly one controlling definitionNo duplicate dispatch or unowned interval between definitions

    Test the actual trigger type, not just the timezone dropdown. Microsoft’s Logic Apps recurrence guidance distinguishes built-in scheduling from managed triggers whose later runs can depend on execution history and other factors. It also discusses recurrence drift. A behavior documented for one trigger is not proof for another.

    WeWeb’s recurring-schedule documentation, for example, describes wall-clock scheduling that skips a nonexistent time and runs once when a time repeats. Use current documentation to establish expected engine behavior, then test your own configuration and downstream handoff. Do not turn an engine-specific example into a universal cron recipe.

    Release through named states and accepted handoffs

    Use Proposed → Rule approved → Configuration verified → Ready for activation → Observing → Accepted for the series-release case. Any failed gate moves it to Held, with a resolution owner, review time and explicit return state. A rejected proposal closes with a reason; it does not activate a different rule by default.

    Business owner to operator

    The business owner sends an approved contract, intended occurrence set, effective boundary and exception authority. The operator accepts only after confirming the supported trigger, timezone interpretation, test cases and stop conditions. Missing cadence or conflicting instructions keep the case Held; they are not configuration choices for the operator to improvise.

    Operator to receiving team

    The operator supplies the series and occurrence IDs, intended period, resolved start instant, acceptance criteria, actual receiving record and next owner. The receiving owner accepts by confirming access, understanding the period and owning the next action. Until then, the release coordinator retains responsibility for the incomplete handoff.

    Activation is not final acceptance. During the agreed observation window, reconcile expected occurrences with dispatch results, receiving records and any authorized omissions or holds. Close the release only when every expected item is accounted for or an explicit residual exception has an accepted owner and follow-up.

    Keep exceptions visible instead of fixing the clock blindly

    • Unknown zone or missing rule: the business owner resolves the governing location and policy. Hold activation; resume only with an approved contract and renewed tests.
    • Engine behavior conflicts with the obligation: the operator documents the mismatch and supported alternatives. Escalate to the business owner; do not silently convert a required daily review into an authorized skip.
    • Dispatch outcome is uncertain: retain Reconciling as an exception under Observing. The operator checks the original attempt and receiving record before another action; resume when the actual outcome is established.
    • Receiver unavailable or notification fails: the coordinator obtains acceptance from an authorized backup. A sent message does not prove accepted ownership.
    • Rules or timezone data change: the operator identifies affected future occurrences and requests an impact review. Preserve past timestamps and outcomes; do not rewrite history to fit the new interpretation.
    • Work was genuinely missed: retain the original period and route it through the missed recurring-task recovery workflow. A new schedule is not retrospective completion.

    If the change concerns counted service hours or existing SLA commitments rather than recurring task creation, use the separate SLA calendar change-control model.

    A worked review: one missing start, one repeated start

    Illustrative design example: an internal queue-review series normally starts at 02:30 in a zone whose spring clock change skips that time. The business requires a review on that date. Its owner approves 03:30 as that date’s exception, with the same business-date occurrence identity and unchanged evidence requirement. The operator must prove that the selected configuration can deliver this result; if it only skips, the release stays Held for a supported alternative.

    A second series starts at 01:30 in a zone where that time repeats in autumn. The owner requires one daily review, not two. The contract chooses the later resolved instant. Tests must prove both the selected instant and the single receiving record; observing “one run” without knowing which instant is incomplete evidence.

    Neither decision follows automatically from the phrase “handle daylight saving.” The owner chooses the obligation, the operator proves the implementation, and the receiver accepts the work. These are design examples, not customer results or instructions to alter a live schedule without its owner’s authorization.

    Measure coverage, not just trigger success

    For a fixed window, reconcile expected business occurrences to accepted receiving records, explicitly authorized omissions and unresolved exceptions. State whether a gap-day omission is excluded by the approved series rule or remains an expected obligation awaiting disposition. Never drop it from the denominator after the fact.

    • Dispatch coverage: expected occurrences with a confirmed receiving record divided by expected occurrences requiring dispatch.
    • Start-time conformance: confirmed starts within the contract’s permitted window; retain intended and actual timestamps.
    • Duplicate effects: unintended additional receiving tasks or consequential actions, separate from harmless logged retries.
    • Exception age and handoff delay: unresolved time by accountable owner, with blocked reasons.
    • Accepted work completion: evidence accepted for the intended period, reported separately from successful dispatch.

    Review these around the relevant transition and after configuration changes. Use recurring failures to improve the contract, capacity or implementation—not to normalize manual timestamp edits.

    Where Workhint fits

    Workhint’s Workflow Engine describes role-based approval routing and requirements for details, files or notes before work progresses. Those capabilities are relevant to approving a clock contract and accepting configuration evidence.

    The workflow automation software solution describes configurable forms, stages, owners, permissions and notifications. That makes it relevant to coordinating the release review and receiving-team handoff. These pages do not establish exact timezone scheduling, gap/fold selection, occurrence uniqueness or historical audit-retention guarantees. Verify those separately in the actual configuration; keep the authoritative timer in an already-supported system if necessary.

    Build This in Workhint

    Use this proposed specification for configuration review and acceptance testing, not as proof that every requirement is implemented:

    Coordinate the release of recurring internal reviews. Give business owners approval rights, operators configuration-evidence tasks and receiving owners access to their handoffs. Intake captures series ID, cadence/anchor, named timezone, period covered, gap/fold policy, start window, overlap/outage treatment and effective boundary. Route Proposed through Rule approved, Configuration verified, Ready for activation, Observing and Accepted. Hold missing inputs, unsupported behavior, conflicts and failed handoffs with named owners and resume gates. Require approved policy, expected occurrence sets, test evidence and receiving acceptance. Notify owners of review requests, rejected tests, unresolved outcomes and overdue handoffs. Provide views for pending approvals, held releases, observed occurrences and residual exceptions. Measure dispatch coverage, time conformance, duplicates and acceptance delay. Require actor, time, reason, rule version and evidence links for transition history, and verify the actual retention and permissions before rollout.

    Do not activate automatic occurrence creation or claim duplicate prevention until those mechanisms are demonstrated. If they remain outside the verified setup, coordinate the review here and retain scheduling and reconciliation in the supported existing system.

    Frequently asked questions

    Should every recurring task use UTC?

    No. Choose whether the obligation follows a local clock or an elapsed interval. Record resolved instants in UTC where useful, but preserve the business timezone and rule that explain them.

    Does one scheduler invocation prove there was no duplicate work?

    No. Check the receiving record and consequential actions. Delivery retries or a second active schedule may still produce unintended duplicate effects.

    Can we simply run a skipped task the next morning?

    Only under the applicable recovery authority. Preserve its original period and lateness; do not label a new morning’s work as evidence for a missed obligation it did not actually cover.

    What if the platform cannot implement the approved clock rule?

    Hold activation and choose a supported alternative with the business owner. Do not silently replace the obligation with whatever the scheduler happens to do.

    Comments

    Leave a Reply

    Your email address will not be published. Required fields are marked *


    The reCAPTCHA verification period has expired. Please reload the page.