B3: Fix tfact_enrollment incremental watermark to capture status updates

[quazi-h]

What are the relevant tickets?

Closes #2123
Epic: #2073

Description (What does it do?)

The incremental filter for tfact_enrollment used only enrollment_created_on as the watermark for platforms that have no updated_on column (edxorg, residential). This silently missed status changes such as enrollment deactivations (is_active: true → false) and mode upgrades (audit → verified).

Platforms that already expose an updated_on timestamp (MITxOnline, MITxPro, program enrollments) use the existing coalesce(enrollment_updated_on, enrollment_created_on) watermark and are unaffected.

The fix adds a targeted 7-day lookback clause scoped to rows where enrollment_updated_on is null:

-- For platforms without updated_on (edxorg, residential), apply a 7-day lookback to catch
-- status changes (deactivations, mode upgrades) that don't bump enrollment_created_on
or (ewf.enrollment_updated_on is null and ewf.enrollment_created_on >= current_timestamp - interval '7' day)

Source audit (confirmed via OpenMetadata):

• raw__mitx__openedx__mysql__student_courseenrollment (residential) — columns: id, mode, created, user_id, course_id, is_active — no updated column in source
• stg__edxorg__bigquery__mitx_person_course (edxorg) — BigQuery snapshot, no updated_on

How can this be tested?

Run the model incrementally against production and verify that deactivated enrollments for edxorg/residential are correctly captured:

uv run dbt run --select tfact_enrollment --vars 'schema_suffix: qhoque' --target dev_production

Then spot-check that a known deactivated residential or edxorg enrollment (where enrollment_is_active = false and enrollment_created_on predates the previous watermark) now appears in the output:

select platform, enrollment_is_active, enrollment_mode, enrollment_created_on, enrollment_updated_on
from ol_warehouse_production_qhoque_dimensional.tfact_enrollment
where platform in ('edxorg', 'residential')
  and enrollment_is_active = false
order by enrollment_created_on desc
limit 20;

Additional Context

tfact_enrollment is a leaf node with no downstream dbt models (confirmed via OpenMetadata lineage). The delete+insert incremental strategy with unique_key='enrollment_key' safely handles re-processing of rows within the lookback window.

[Copilot]

Pull request overview

Adjusts tfact_enrollment’s incremental filter so platforms that lack an updated_on source field (edxorg, residential) still reprocess a small recent window and don’t miss enrollment status/mode changes.

Changes:

• Adds a 7-day lookback condition for rows where enrollment_updated_on is null in the incremental WHERE clause.
• Keeps existing coalesce(enrollment_updated_on, enrollment_created_on) watermark behavior for platforms that do have updated_on.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated 1 comment.

[KatelynGit]

build ran clean and it does appear to be picking up additional records now