diff --git a/.github/POSTGRES_COMPATIBILITY.md b/.github/POSTGRES_COMPATIBILITY.md
index b4b7afbb5a3..13378137e59 100644
--- a/.github/POSTGRES_COMPATIBILITY.md
+++ b/.github/POSTGRES_COMPATIBILITY.md
@@ -180,6 +180,13 @@ audit of these fixes found four recurring mistakes:
   the arbitrary-pick preservation the wrap is usually justified as. Confirmed on CI; see #56241.
   Note a local macOS PostgreSQL gives a **false all-clear** — its collation happens to agree with
   MariaDB on case. Fix: take a representative row rather than sorting text.
+  **Picking that row is the hard part.** `Min(name)` is still a text sort: `autoname="hash"` is
+  not reliably lower case, because `_get_timestamp_prefix()` prepends `get_trace_id()[-1:]`
+  un-lowered and a client-supplied `X-Frappe-Request-Id` can put an upper case `A-F` there. A
+  non-text key (`Min(idx)`) works only where it is **unique within the group** and the join-back
+  carries the **full group key** — a date is usually neither, and joining on a duplicated value
+  turns one group into several rows (§3). Otherwise select the row in Python, sorting with
+  `key=str.casefold` so the order matches MariaDB's collation without depending on the database's.
 - **Wrong bound** — where the value has a semantic, pick the bound deliberately:
   `Min(schedule_date)` for a "required by", `Min(idx)` for first-line ordering, a qty-weighted
   average for a rate. A blind `Max` can understate urgency or overstate a figure.
diff --git a/erpnext/accounts/utils.py b/erpnext/accounts/utils.py
index 09b5a60fc53..2837ed47a28 100644
--- a/erpnext/accounts/utils.py
+++ b/erpnext/accounts/utils.py
@@ -2415,8 +2415,8 @@ class QueryPaymentLedger:
 			.groupby(ple.account, ple.voucher_type, ple.voucher_no, ple.party_type, ple.party)
 		).as_("grouped")
 
-		# Payment Ledger Entry has no autoname rule, so frappe names it by hash -- lower-case, which
-		# keeps Min(name) free of the collation divergence that picking Max() over free text has.
+		# KNOWN DIVERGENCE: Min(name) is a text sort. Hash names are not reliably lower case -- the
+		# trace-id prefix is not lowered -- so the engines can pick different rows here.
 		representative_ple = qb.DocType("Payment Ledger Entry").as_("representative_ple")
 		query_voucher_amount = (
 			qb.from_(grouped_voucher_amount)