CHANGELOG: the brains-review wave, one splice (17 entries)

[erikdarlingdata]

The one post-wave changelog splice per the buffered protocol: wave-1's sixteen PRs merged to dev without touching CHANGELOG.md; this PR adds the seventeen Fixed entries (#3523-#3537, #3547, #3548, #3551, #3556) under a fresh [Unreleased] section plus reference links. #3549 was doc-only, no entry. changelog_archive.py verify passes (all archive shas intact).

🤖 Generated with Claude Code

https://claude.ai/code/session_01QMYFB4qv5tqGwMmdPsa4Zx

[claude]

Reviewed. This is a documentation-only change (CHANGELOG.md, no T-SQL/C#), so the Lite/Darling parity and T-SQL style checks don't apply. Structurally the diff checks out: CRLF preserved, all 20 new [#N] reference links resolve exactly once with no collisions against the pre-existing definition block, and none of the 20 issue numbers were already documented elsewhere in the file.

One thing worth a look before merge — the PR description's accounting doesn't match the diff:

• It says "the seventeen Fixed entries (get_top_queries_by_cpu and get_top_procedures_by_cpu rank by elapsed time, not CPU #3523-The xmin horizon alert false-fires on 1-of-1 observations and misses rotating holders #3537, QueryStore slicer plots logical reads under the "Avg Physical Reads" label #3547, get_memory_trend: join the grants series so total_granted_mb carries real data (complete fix for #3529) #3548, Viewer Recommendations tabs still say "All clear" when the analysis window collected nothing (viewer half of #3524) #3551, Procedures grid plots logical reads under the physical-reads label, and two slicer labels call totals averages (#3547 siblings) #3556)" and calls out only Two stale doc lines still describe get_memory_trend's total_granted_mb as an always-0 parity placeholder #3549 as a doc-only gap in that range.
• The actual diff adds 20 entries, not 17, and two of them — Deprecated Dashboard analysis has the #3527 perfmon divisor defect (per-interval delta read as per-second) #3561 and Viewer pass for #3528: Settings box for the Store Disk Pressure GB floor, and measured-metric qualifier on the WPF fleet surfaces #3563 — aren't mentioned anywhere in the description.
• Also, Anomaly findings max at severity 1.0 under a 1.5 notify default, so the baseline engine can never page at shipped settings #3526 falls inside the stated #3523-#3537 range but has no entry and isn't called out the way Two stale doc lines still describe get_memory_trend's total_granted_mb as an always-0 parity placeholder #3549 is, so it's unclear if that's an intentional second gap or an oversight in the write-up.

Since the buffered-changelog protocol relies on this description being an accurate manifest of what got spliced in, it'd be good to confirm #3561/#3563 were meant to be included in this wave (vs. leaking in from a different batch) and to note the #3526 gap explicitly, the same way #3549 is noted.

[claude]

LGTM — doc-only CHANGELOG.md splice; verified CRLF preserved, all 20 new [#N] links resolve with no duplicates/collisions, no code touched so Lite/Darling parity and T-SQL style don't apply. Left a comment noting the PR description's entry count/list doesn't match the actual diff (17 claimed vs 20 added, #3561/#3563 unmentioned, #3526 gap unexplained) — non-blocking, description accuracy only.