For reference and from practice.

16 sessions · queries per second
 └ 4 cores, baseline             163
 └ 8 cores, otherwise unchanged  234
 └ 4 cores, query improved       584

Same code · CL 160
 └ CPU on 2022            107–109 ms
 └ CPU on 2025            87–89 ms
engine ≠ level ≠ feature

Ola · LOG run on SIMPLE
 └ return code            0
 └ new LOG backup sets    0
execution ≠ coverage

Restore · 3 rows vs. 3 rows
 └ STOPAT before DELETE   1, 2, 3
 └ latest log state       1, 2, 4
row count ≠ data state

Lead or cause?
 └ Index · Present            Prio 1
 └ Parameter · CompiledValue  Prio 10
cross-check, reads  270,158 / 2,790

no maintenance, 3.6 million rows
 └ page density           ≈ 68% stable
 └ per 10,000 inserts     ≈ 225 split log records
a quiet index?            no

100,000 inserts · split log records
 └ after REORGANIZE       14,870
 └ after REBUILD FF81     2,632
fragmentation before      0.28 % · 0 %

target page [10 20 30 40]  ← 25?
 └ fragmentation          0 %
 └ free slots             0
next row                  split

EXEC dbo.p2 800;   -- READ COMMITTED
 └ S key locks ix_ID_Tab1  6,220
 └ two rows later          S on table
INSERT waits               LCK_M_IX

UPDATE dbo.Control       7,687.8 ms
 └ Trigger → Procedure
   └ UPDATE dbo.WorkTarget 7,666.6 ms
Σ 15,354 ms · elapsed    7,688 ms

WHERE           --#if @Kunde IS NOT NULL
(               --#-
@Kunde IS NULL  --#-
 OR k.ID=@Kunde --#if @Kunde IS NOT NULL
)               --#-

   0 m  SELECT * · NOLOCK
 110 m  Parameter Sniffing
 320 m  Spinlocks · VLFs
 448 m  Cthulhu