PerformanceMonitor never runs the plan BenefitScorer, so findings get no benefit score and the scorer's wait findings never appear #4546

Claude posting for Erik Darling

PerformanceMonitor never runs BenefitScorer, so no plan finding gets a benefit score, and the scorer's own findings never appear.

Seen

BenefitScorer.Score(ParsedPlan) (PerformanceMonitor.PlanAnalysis/BenefitScorer.cs) exists, and its own doc comment says it "runs after PlanAnalyzer.Analyze()". But no product code calls it. git grep -n "BenefitScorer.Score" -- '*.cs' ':!*Tests*' finds nothing. Every entry point runs only the analyzer:

So PlanWarning.MaxBenefitPercent and PlanStatement.WaitBenefits are always empty. Nothing reads them in the viewer, MCP output or Lite either.

Why it matters now

The plan-analysis sync (#4511) ports PerformanceStudio changes that live in the scorer, and none of them has any effect in PM until the scorer runs:

• the wait-stat findings (Wait stats never become findings: BenefitScorer computes a benefit for each wait but emits no warning #4516), which the scorer emits;
• the external/preemptive wait benefit (The benefit for external and preemptive waits is far too low (MEMORY_ALLOCATION_EXT, PREEMPTIVE_*) #4517);
• "Bare Scan" joining the operator-time benefit rules (Add rule 34: a bare scan with no predicate, with advice for a covering index or columnstore #4526);
• the Serial Plan benefit formula (Serial Plan (rule 3) benefit uses the older two-branch formula #4528).

PerformanceStudio runs both steps in one place (PlanAnalysisPipeline: analyze, then score).

Fix shape

Run BenefitScorer.Score(plan) right after PlanAnalyzer.Analyze(plan) at each entry point above, or through one shared analyze-then-score helper like PerformanceStudio's. Then decide whether the viewer and the MCP output show each finding's benefit. The deep-plan stack fix (#4512) covers the scorer's recursion too.