Which connector?: 

Connector
 ukg_pro_workforce_management
 Connection: workforce (signal_corps)
 Destination: Snowflake UKG_PRO_REPLICATION.WORKFORCE

Additional details:

Problem
 The Retrieve Timecards as Manager API (POST /v1/timekeeping/timecard/multi_read, select: ACTIVITY_SHIFTS) returns current activity state only. When an employee’s activity is edited in UKG, prior activities (commonly IDLE) disappear from the payload. They are not returned as deleted records.

Fivetran currently does not support delete capture for TIME_CARD_AS_MANAGER and related child tables, including TIME_CARD_AS_MANAGER_ACTIVITY_SHIFT_WORKED_SPAN. Rows that were synced earlier stay in the warehouse with _fivetran_deleted = false after they are gone from UKG. Support confirmed there is no re-extract path that will soft-delete them today.

Expected behavior
 On each sync, reconcile nested activity rows against the current multi_read payload for that employee and date range. If a previously synced child row (for example grain employee_id + work_shift_id + index, or the processed-segment id) is absent from the current payload, set _fivetran_deleted = true on the next sync. Same treatment for other TIME_CARD_AS_MANAGER child tables sourced from this endpoint.

Why this matters
 Idle hours and other dropped activities keep flowing into downstream labor reporting because models correctly filter _fivetran_deleted = false. That overstates idle versus the live UKG UI and versus a current API call. Payroll totals are not the issue. Activity allocation is. A historical refresh does not fix it, which we already verified.

Reproducible example
 Employee qualifier 126036 (Snowflake EMPLOYEE.ID 52161), date 2026-07-15, WORK_SHIFT_ID 5158783.

Current UKG payload (confirmed by UKG, Brian Harrington) contains only:

• CO/WIRE/131107-27 (processed segment 6222653, 08:00-16:30)
• MEALBREAKUNPAID (processed segment 6222265, 12:00-12:30)

No IDLE.

Fivetran still has three rows in TIME_CARD_AS_MANAGER_ACTIVITY_SHIFT_WORKED_SPAN:

• index 0, CO/WIRE, _fivetran_synced 2026-09-07, _fivetran_deleted false
• index 1, MEALBREAKUNPAID, _fivetran_synced 2026-09-07, _fivetran_deleted false
• index 2, IDLE 4 hours (12:30-16:30, processed segment 6222266), _fivetran_synced 2026-07-15, _fivetran_deleted false

Indexes 0 and 1 were upserted on a later sync. Index 2 was left in place.

Support ticket: Meghana Bale confirmed this is current connector behavior (no delete capture, no absence-based reconciliation for these nested records).

Ask
 Add delete capture / current-state reconciliation for TIME_CARD_AS_MANAGER nested activity tables so warehouse state matches the current UKG timecard payload.