Plan değişimi P401

(@ServerID int,@DatabaseName nvarchar(128))INSERT perf.QdsRuntimeStats
        SELECT g.* FROM perf.Stg_QdsRuntimeStats AS g
        WHERE g.ServerID = @ServerID AND g.DatabaseName = @DatabaseName
          AND NOT EXISTS (SELECT 1 FROM perf.QdsRuntimeStats AS r
                          WHERE r.ServerID = g.ServerID AND r.DatabaseName = g.DatabaseName AND r.plan_id = g.plan_id
                            AND r.runtime_stats_interval_id = g.runtime_stats_interval_id AND r.execution_type = g.execution_type
                            AND r.IntervalStartUtc = g.IntervalStartUtc)

Plan change case #P401 | DB: DB-12 | Improvement | Confidence: Low (review pending)
Query: (@ServerID int,@DatabaseName nvarchar(?))INSERT perf.QdsRuntimeStats SELECT g.* FROM perf.Stg_QdsRuntimeStats AS g WHERE g.ServerID = @ServerID AND g.DatabaseName = @DatabaseName AND NOT EXISTS (SELECT ? FROM perf.QdsRuntimeStats AS r WHERE r.ServerID = g.ServerID AND r.DatabaseName = g.DatabaseName AND r.plan_id = g.plan_id AND r.runtime_stats_interval_id = g.runtime_stats_interval_id AND r.execution_type = g.execution_type AND r.IntervalStartUtc = g.IntervalStartUtc)
Change: 2026-09-28 13:00 UTC, plan 4448 → 5955, pattern: Alternating (27 switches)
Impact (Query Store, per execution): duration 110.4 → 6 ms (-95%); CPU 81.4 → 4.6 ms (-94%); reads 14,863 → 1,056 (-93%); rows 2117.18 → 96.13 (-95%)
Plan difference: Join only in the old plan: Hash Match (Left Anti Semi Join); Join only in the new plan: Nested Loops (Left Anti Semi Join); Estimated cost: 0.902 → 0.379; estimated rows: 1 → 1
Candidate cause: WORKLOAD — The plans see a different workload: rows per execution 2117.18 → 96.13. The plans may have been compiled for different parameters or batch sizes.
Candidate cause: PARAM — The two plans alternate in the same period (27 switches): plan choice depends on the parameter value rather than an event (parameter sensitivity).
Counter-evidence: WORK_PROPORTIONAL — Rows per execution changed 0.05× and duration 0.05×; duration per row 0.0522 → 0.0625 ms: the duration difference comes from the amount of data processed, not the plan.
Counter-evidence: HIGH_VARIANCE — The old plan's duration standard deviation is 165% of the average: the average is not reliable.
Counter-evidence: LOW_SAMPLE — The old plan ran only 28 times: the sample is small and the average is weak.
Counter-evidence: HIGH_VARIANCE — The new plan's duration standard deviation is 253% of the average: the average is not reliable.
Counter-evidence: METADATA_GAP — Metadata collection started on 2026-10-01 and the change was on 2026-09-28: index, statistics and module events could not be evaluated (no data, not no events).
Confidence rationale: Low: strong counter-evidence (WORK_PROPORTIONAL).
Note: this is a candidate cause, not proof of the root cause.