docs(series4): the 2026-09-22 failure is accounted for -- the timeline was wrong

An earlier draft said the Prolific-cable finding did not fit the timeline, since
the unit was described as working and then degrading, which a cable that never
enumerates cannot do.  The timeline was what needed correcting.

From the operator: the initial setup was ONE modem and ONE cable, and it worked.
Two Micromates then had to be deployed, so a second cable was fetched to run both
at once -- and only one of the two ever worked.  The other never talked over its
modem at any point, was never deployed, and a MiniMate Plus went out in its place.
That Micromate came back to the bench, and the cable on it is a Benfei (Prolific
PL2303).

So nothing degraded.  The "worked at first" was the single-cable setup before the
split.  One unit working and one not, configured side by side, is the signature of
a mixed cable supply rather than a unit fault.

Still worth confirming whether that cable travelled home with the unit or was
picked up at the bench; if it travelled with it there is no inference left.  The
FTDI cable arriving 2026-09-26 settles it either way -- same unit, same modem, one
variable.

Records a corollary: "just deploy a Series III instead" works partly because a
MiniMate Plus has a DB-9 port directly on the unit, with no USB-to-serial adapter
in the path and therefore no chipset to get wrong.  That workaround has been
routing around this exact failure mode.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ru8Lg9HkkYvX9VWWo65SmL

docs/micromate_protocol_reference.md

@@ -2119,20 +2119,40 @@ FTDI is VID `0403`, Prolific is `067b`.  Note also that counterfeit FTDI chips
 are common in cheap cables — they carry FTDI's VID but may not behave like one,
 and an embedded host with a single driver is far less forgiving than Linux.
 
-### Does this explain the 2026-09-22 field outage?
+### ✅ This accounts for the 2026-09-22 failure
 
-⚠ **Not on the timeline as reported.**  A Prolific cable does not fail
-*gradually* — it never enumerates, so the modem port never works at all.  The
-unit was described as working at first and degrading.
+An earlier draft of this section said the timeline did not fit, because the unit
+was described as *working at first and then degrading* — which a Prolific cable
+cannot do, since it never enumerates at all.
 
-There is one story where it fits: if the initial success was over a **different
-path** — the USB **PC** port, or a bench test before deployment — that would work
-regardless of which serial cable was attached.  The modem path would then have
-been broken from the moment it was deployed, and "it worked and then stopped"
-would be a recollection conflating two connection types days after the fact.
+**The timeline was the thing that needed correcting, not the finding.**  From the
+operator, 2026-09-25:
 
-Plausible, unverified, and recorded as such.  Identifying that unit's cable would
-settle it.
+1. Initial setup — **one** modem, **one** cable.  Worked: configs sent, status
+   read, no trouble.
+2. Two Micromates then had to be deployed, so a **second cable** was fetched to
+   run both at once.
+3. **Only one of the two ever worked.**  The other would not talk over its modem
+   at any point.
+4. That unit was never deployed — a **MiniMate Plus was put out in its place**,
+   and the Micromate came back to the bench.
+5. The cable on that bench unit is a **Benfei (Prolific PL2303)**.
+
+So nothing degraded.  The "worked at first" was the original single-cable setup;
+the failure began when the pair was split across two cables and one of them was
+the wrong chipset.  **One working and one not, set up side by side, is the
+signature of a mixed cable supply** — not of a unit fault.
+
+⚠ Still worth confirming: whether that cable travelled home *with* the unit, or
+was picked up separately at the bench.  If it travelled with it, there is no
+inference left.  Either way the FTDI cable arriving 2026-09-26 settles it — same
+unit, same modem, one variable.
+
+**Corollary worth knowing.**  The standing workaround — *"just deploy a Series III
+instead"* — works partly because a **MiniMate Plus has a DB-9 port directly on the
+unit.**  No USB-to-serial adapter anywhere in the path, so there is no chipset to
+get wrong.  That workaround has been quietly routing around this exact failure
+mode.
 
 ### How this was isolated — the method is reusable