docs: expand Awards wiki — Test/Explain, Missing refs, n-fer, catalog updates

The matching "five rules" were already documented; add the parts that were only
in code: the per-QSO Explain trace (Test a callsign in the award editor), the
Missing-refs gap finder (in-scope-but-no-reference + bulk assign), how one QSO
credits several references (n-fer POTA / VUCC grid lines, comma/semicolon split),
code-vs-exact matching, and how built-in award updates skip awards you've edited.

wiki/Awards.md

@@ -23,6 +23,17 @@ Plus a **leading/trailing** strip, a **prefix** prepended to found references, a
 **dynamic** mode (any value counts, like POTA), and the **fallback searches**
 described below.
 
+**One QSO can count for several references.** A contact always yields *every*
+reference its field holds — the value is split on commas and semicolons, so an
+n-fer POTA activation logged as `US-6544,US-0680` credits both parks, and a
+grid-line VUCC contact credits each grid. There is no "one reference per QSO"
+switch — it was never useful.
+
+**Match by `code`** looks up each token of the field in the reference list (so a
+partial or multi-value field still resolves), while **`exact match`** requires
+the whole field to equal the reference. Use exact match when a substring would
+cause false hits.
+
 ### The five rules to remember
 
 1. **Every regex is case-insensitive.** Award and per-reference patterns match
@@ -72,6 +83,42 @@ A custom province award where the log rarely spells the province out (it names a
 Now a QSO is resolved by the province **name** if present, else by a **city**
 regex — first in the QTH, then in the address. First hit wins.
 
+## Why doesn't a QSO count? — Test & Missing refs
+
+An award that matches nothing is the hardest thing to debug: the column is just
+empty, and you can't tell whether the QSO was out of scope, the field was empty,
+the rule looked in the wrong place, or the reference isn't on the list. Two tools
+answer that.
+
+### Test a callsign (award editor)
+
+In the **award editor** there's a **Test** box: type a callsign and OpsLog
+replays that award's rules against every QSO you have with that station and shows
+what happened, step by step — the exact same code path the real matcher uses, so
+what you see is what actually runs. For each contact it reports:
+
+- whether the QSO is **in scope** (and, if not, *why* — wrong DXCC, band, mode,
+  emission, or outside the valid-from/to dates);
+- each rule in turn (**primary**, then **Fallback 1, 2, …**), the **field value**
+  it scanned, the **candidates** it produced, which ones were **kept**, and which
+  were **rejected** (with the reason — e.g. "not on the reference list");
+- rules that were **skipped** because an earlier one already matched;
+- any **manual** reference you assigned by hand;
+- the final **result** — what the QSO counts for.
+
+This is the fastest way to fix a rule: if the primary scanned the wrong field, or
+a candidate was rejected as unlisted, the trace says so directly.
+
+### Missing refs (awards panel)
+
+For an award scoped to a DXCC entity (DDFM, WAS, RAC, WAJA…), the **Missing refs**
+view lists contacts that ARE in scope (right DXCC / band / mode / dates) **but
+where no reference was found** — so they don't count yet. Sort by a column, tick
+the matching contacts, and **assign the reference** to all of them at once (or
+click a row to open the QSO and fix the field). Recompute and the fixed contacts
+drop off the list. This is how you close the gaps a matcher can't fill on its own
+(a French QSO logged without its department, say).
+
 ## Live detection & manual refs
 
 - References are detected **live** as you enter a callsign.
@@ -89,6 +136,15 @@ regex — first in the QTH, then in the address. First hit wins.
   or **both**. Award columns are opt-in per the Columns picker
   ([[Recent QSOs and Filters]]).
 
+## Built-in updates vs your edits
+
+Built-in awards ship with the app, definition **and** reference list. When a
+newer version fixes a shipped award (a better fallback chain, a corrected
+reference list), OpsLog offers to apply that update — **unless you've edited that
+award yourself**, in which case your version is left alone (your work outranks
+ours). Awards you created are never touched by updates. So you can freely tune a
+built-in award without fear of a future release overwriting it.
+
 ## Rescan
 
 **Rescan** re-pulls the logbook and recomputes — it picks up fresh LoTW / QRZ /