Plan değişimi P445

(@incTest bit,@types1 nvarchar(4000),@sid int,@dbName nvarchar(4000),@tag nvarchar(4000),@model nvarchar(4000),@q vector(1024),@top int)WITH x AS (
    SELECT c.ChunkId, d.DocId, d.SourceType, d.SourceRef, d.Title, c.Heading, c.Text,
           VECTOR_DISTANCE('cosine', c.Embedding, @q) AS Score,
           ROW_NUMBER() OVER (PARTITION BY d.DocId, c.Seq ORDER BY VECTOR_DISTANCE('cosine', c.Embedding, @q), c.ChunkId) AS rn,
           MIN(IIF(o.PairDocId IS NULL, c.ChunkId, NULL)) OVER (PARTITION BY d.DocId, c.Seq) AS OwnChunkId
    FROM rag.Chunk AS c
JOIN rag.Document AS o ON o.DocId = c.DocId
JOIN rag.Document AS d ON d.DocId = ISNULL(o.PairDocId, o.DocId)
    WHERE (d.IsActive = 1 OR (@incTest = 1 AND d.ExclusionReason = 'TestDb')) AND d.SourceType IN (@types1)
AND (@sid IS NULL OR d.ServerID = @sid) AND (@dbName IS NULL OR d.DatabaseName = @dbName)
AND (@tag IS NULL OR EXISTS (SELECT 1 FROM STRING_SPLIT(d.Tags, ',') AS t WHERE LTRIM(RTRIM(t.value)) = @tag)) AND c.Embedding IS NOT NULL AND c.EmbeddingModel = @model
)
SELECT TOP (@top) ISNULL(e.ChunkId, x.ChunkId) AS ChunkId, x.DocId, x.SourceType, x.SourceRef, x.Title,
       IIF(e.ChunkId IS NULL, x.Heading, e.Heading) AS Heading, ISNULL(e.Text, x.Text) AS Text, x.Score
FROM x
LEFT JOIN rag.Chunk AS e ON e.ChunkId = x.OwnChunkId
WHERE x.rn = 1
ORDER BY x.Score, ISNULL(e.ChunkId, x.ChunkId)

Plan change case #P445 | DB: DB-11 | Regression | Confidence: Low (review pending)
Query: (@incTest bit,@types1 nvarchar(?),@sid int,@dbName nvarchar(?),@tag nvarchar(?),@model nvarchar(?),@q vector(?),@top int)WITH x AS ( SELECT c.ChunkId, d.DocId, d.SourceType, d.SourceRef, d.Title, c.Heading, c.Text, VECTOR_DISTANCE(?, c.Embedding, @q) AS Score, ROW_NUMBER() OVER (PARTITION BY d.DocId, c.Seq ORDER BY VECTOR_DISTANCE(?, c.Embedding, @q), c.ChunkId) AS rn, MIN(IIF(o.PairDocId IS NULL, c.ChunkId, NULL)) OVER (PARTITION BY d.DocId, c.Seq) AS OwnChunkId FROM rag.Chunk AS c JOIN rag.Document AS o ON o.DocId = c.DocId JOIN rag.Document AS d ON d.DocId = ISNULL(o.PairDocId, o.DocId) WHERE (d.IsActive = ? OR (@incTest = ? AND d.ExclusionReason = ?)) AND d.SourceType IN (@types1) AND (@sid IS NULL OR d.ServerID = @sid) AND (@dbName IS NULL OR d.DatabaseName = @dbName) AND (@tag IS NULL OR EXISTS (SELECT ? FROM STRING_SPLIT(d.Tags, ?) AS t WHERE LTRIM(RTRIM(t.value)) = @tag)) AND c.Embedding IS NOT NULL AND c.EmbeddingModel = @model ) SELECT TOP (@top) ISNULL(e.ChunkId, x.ChunkId) AS ChunkId, x.DocId, x.SourceType, x.SourceRef, x.Title, IIF(e.ChunkId IS NULL, x.Heading, e.Heading) AS Heading, ISNULL(e.Text, x.Text) AS Text, x.Score FROM x LEFT JOIN rag.Chunk AS e ON e.ChunkId = x.OwnChunkId WHERE x.rn = ? ORDER BY x.Score, ISNULL(e.ChunkId, x.ChunkId)
Change: 2026-10-07 18:00 UTC, plan 65863 → 65864, pattern: Alternating (10 switches)
Impact (Query Store, per execution): duration 18.6 → 49.4 ms (+166%); CPU 18.5 → 49.3 ms (+167%); reads 5,328 → 20,609 (+287%); rows 12.15 → 17.58 (+45%)
Plan difference: no operator-level difference
Candidate cause: STATS — Statistic rag.Chunk.PK_Chunk was updated 2026-10-07 16:19 (sampled 39%); 3 updates in the window; the plans use this statistic.
Candidate cause: STATS — Statistic rag.Chunk.UQ_Chunk_DocSeq was updated 2026-10-07 16:19 (sampled 55.2%); 3 updates in the window; the plans use this statistic.
Candidate cause: STATS — Statistic rag.Chunk._WA_Sys_00000007_49309088 was updated 2026-10-07 16:19 (sampled 54.7%); 3 updates in the window; the plans use this statistic.
Candidate cause: STATS — Statistic rag.Chunk._WA_Sys_00000003_49309088 was updated 2026-10-07 16:19 (sampled 39%); 3 updates in the window; the plans use this statistic.
Candidate cause: STATS — Statistic rag.Document.PK_Document was updated 2026-10-07 10:27 (sampled 100%); 2 updates in the window; the plans use this statistic.
Candidate cause: STATS — Statistic rag.Document.UQ_Document_Source was updated 2026-10-07 13:59 (sampled 100%); 2 updates in the window; the plans use this statistic.
Candidate cause: STATS — Statistic rag.Document._WA_Sys_0000000A_428392F9 was updated 2026-10-06 19:19 (sampled 100%); 2 updates in the window; the plans use this statistic.
Candidate cause: STATS — Statistic rag.Document._WA_Sys_00000005_428392F9 was updated 2026-10-07 16:19 (sampled 100%); 3 updates in the window; the plans use this statistic.
Candidate cause: STATS — Statistic rag.Document._WA_Sys_0000000E_428392F9 was updated 2026-10-07 13:59 (sampled 100%); 3 updates in the window; the plans use this statistic.
Candidate cause: STATS — Statistic rag.Document._WA_Sys_00000007_428392F9 was updated 2026-10-07 16:18 (sampled 100%); 2 updates in the window; the plans use this statistic.
Candidate cause: STATS — Statistic rag.Document._WA_Sys_00000010_428392F9 was updated 2026-10-07 16:18 (sampled 100%); the plans use this statistic.
Candidate cause: PARAM — The two plans alternate in the same period (10 switches): plan choice depends on the parameter value rather than an event (parameter sensitivity).
Counter-evidence: INDEX_UNUSED — rag.Document.IX_Document_Pair was created in the window but the new plan does not use it: candidate ruled out.
Counter-evidence: HIGH_VARIANCE — The new plan's duration standard deviation is 155% of the average: the average is not reliable.
Counter-evidence: SAME_SHAPE — The two plans have the same operator tree (operators, tables, indexes): the plan did not change; the difference is only in the compiled parameter values and row estimates. The slowness may come from the amount of data this parameter value processes or from waits, not from plan choice.
Confidence rationale: Low: strong counter-evidence (SAME_SHAPE).
Note: this is a candidate cause, not proof of the root cause.