docs(wiki): cover the recent work, with worked examples

Asked for the wiki to catch up, and to be explicit — examples rather than
descriptions. New page and six rewritten sections.

Recent QSOs and Filters, the page named in the request: the right-click menu is
now a table of what every entry does, and the filter builder has five worked
multi-criteria filters written out field by field, each ending in what you then
do with the result. Including the distinction that prompted it — "export
selected" takes the highlighted rows, "export filtered view" takes every QSO the
filter matched, and after building a filter it is the second one you want. Plus
the traps: ADIF date format, and "is empty" not being "equals N".

Connections (new): the page for a settings section that had none. What OpsLog
listens to and sends, which outbound row another logger actually wants — a
logger asking for "WSJT-X UDP" discards a bare ADIF record — the custom-message
triggers with their placeholder lists, three worked examples, and the port-clash
warning that produced this week's CAT-drop report.

Amplifiers and Switches: the generic HTTP relay in full — per-relay URLs, the
pattern form, {relay}, {relay-1} and {value} with what each one sends — relay
automatic control with a four-relay antenna-switch example, and the bench power
supply including why OpsLog writes only its output.

Rotators: the two new native types (SPID with the Rot1Prog/Rot2Prog choice,
DCU-1), two-rotor setups, and that Test only reads the heading on the types
where it does.

Awards: the US county database, and the difference between the automatic fill
(blanks only) and the right-click re-derive (replaces) — with the Connecticut
planning regions, which is the case that needs it.

QSL Management: QSL_VIA against QSL_SENT_VIA/QSL_RCVD_VIA, why an import shows E
on every Log4OM record, and that V counts as confirmed.

Troubleshooting: the three field reports of the last few days, by symptom.

Every [[link]] checked to resolve.

wiki/Troubleshooting.md

@@ -76,3 +76,40 @@ selectable, with a copy button.
 
 Open an issue (or *Send log to F4BPO*) with: what you did, what happened, your
 radio/model, and the relevant log lines.
+
+## The rig disconnects during digital transmissions
+
+Symptom: with WSJT-X or JTDX set to **Fake It** split, the CAT link drops on
+every over. Fake It shifts the dial for each transmission instead of using
+split, so it changes the radio state twice per cycle.
+
+Check that **no inbound Connections row shares a port with an outbound one**.
+If it does, OpsLog receives what it publishes, and its own frequency broadcast
+can come back as a command to re-tune — a loop that fires on every dial shift.
+The log names the clash at startup. Give one of the two rows another port.
+
+## A motorized antenna does not follow the rig
+
+Look for this line in the log:
+
+    antenna: steppir connected, but TRACKING IS OFF in Settings — it will not
+    follow the rig, and only moves when you tune it by hand
+
+That is not a fault. Tracking is a setting (Settings → Antenna), and with it off
+the antenna only moves when you tune it yourself.
+
+If the antenna genuinely stops answering, the log says so and reconnects on its
+own:
+
+    steppir: status query failed, reconnecting: read status: timed out …
+    steppir: reconnected after 2 failed attempt(s)
+
+## A rotator turns the wrong way
+
+On a **SPID**, check the dialect first: **Rot1Prog** for the azimuth-only
+controllers (RAK, RAU) and **Rot2Prog** for the RAS / BIG-RAS / MD-01 / MD-02.
+They use different frame lengths, and the wrong one is the usual cause of a
+heading that reads correctly while every commanded move goes the long way round.
+
+Test with a **small** move first — from 45° to 50°, not to 180° — so a wrong
+result cannot wrap the coax at the top of the tower.