# Leave integrity review: September 6, 2026

Source: `solidmarkmaster_db (11).sql`, exported September 6, 2026 at 03:35 AM.
The complete dump was imported successfully into a disposable MariaDB 10.4.32
instance on port 33316. The corrected preflight executed against that copy.
These findings describe the supplied snapshot, not a live database inspection.
No production data was changed and the integrity migration was not rerun on the copy.

## Decision

MODIFIED - September 6, 2026: The user deferred historical reconciliation and
reported that the missing employee IDs were either deleted or modified; the
specific cause and replacement identities remain unverified. The user's pasted
phpMyAdmin results confirm the same six missing-employee rows and the same
missing balance group (two approved records, 1.88 days).

Company policy development may proceed independently of this deferred cleanup.
Preserve the affected history and continue reporting these findings. Do not
derive verified opening entitlements from the disputed balances or treat this
deferral as a successful integrity clearance. No automatic repair is justified
by the supplied evidence.

## Missing employee references

All six blocker rows have missing employees; their leave types exist and their
reported durations pass the preflight range check.

| Table | Leave ID | Employee ID | Status |
| --- | --- | --- | --- |
| leave_requests | 33 | 12345678 | rejected |
| leave_requests | 34 | 12345678 | rejected |
| leave_requests | 35 | 1 | rejected |
| employee_leaves | 6 | 12345678 | cancelled |
| employee_leaves | 7 | 12345678 | approved |
| employee_leaves | 8 | 12345678 | approved |

Approved rows 7 and 8 total 1.88 days for type 0 in 2026 and have no corresponding
balance row. They have no request link. Do not infer a link from equal IDs.
HR must establish whether these employee IDs were deleted, changed, or used for
testing, using an employee backup or approved source records. Do not invent an
employee or silently delete/cancel these records.

## Balance discrepancies

| Employee ID | Type ID | Year | Stored usage | Active approved usage | Difference |
| --- | --- | --- | --- | --- | --- |
| 10545 | 8 | 2026 | 1.50 | 1.00 | +0.50 |
| 10159 | 9 | 2026 | 1.69 | 1.19 | +0.50 |

The snapshot does not establish whether these differences are valid historical
usage, manual adjustments, or errors. Review approved history and adjustment
evidence before correcting balances; never reset usage to the calculated sum
solely to clear the report.

## Other results

- Migration marker `2026_09_06` is present, applied at `2026-09-06 01:34:01`.
- Both automatic-link ambiguity reports and `leave_link_review` are empty.
- Twenty-two legacy log rows retain `log_id=0`; `entry_id` is now an automatic
  primary key. Preserve the legacy IDs and deduction history.
- Approved rows 2, 3, 4, 5, and 8 contain fractional days. Fractions alone are not
  invalid; verify attendance/payroll handling against the actual leave evidence.
- The independent synthetic integration suite passed all 29 checks, including
  approval, edits, cancellation, balance reversals, ownership and payroll cutoffs.
  This is service-level validation, not a browser or live login test.

## Newly modified SQL

The preflight retains its original checks and adds dated, read-only checks 10
and 11 for missing-reference details and missing balance groups. Check 6 alone
cannot detect a missing balance because it starts from existing balance rows.

## Deferred reconciliation follow-up

When reconciliation resumes, obtain the verified identity/history for employee IDs `12345678` and `1`, and
the reason for each 0.50-day usage difference. Then prepare an explicit audited
reconciliation, test it on a fresh copy, and rerun the preflight. This information
is needed for historical repair, not to start company policy development.
Policy activation must account for unresolved balances explicitly; development
does not authorize silently repairing or certifying them.