Upstream partial service: bonded to 1 of 4 channels, T3/UCD storms every 11-12 minutes, and a 21-minute registration failure

I have an intermittent but persistent upstream problem that started this morning
and is still going. Downstream looks fine, upstream does not.

I have logged the modem's own diagnostics every 5 minutes since 10:30 today. Requesting a return-path sweep and, if a corporate
moderator can, a CMTS-side look at my node history.

EQUIPMENT
Netgear CM1000, hardware 2.02, firmware V7.01.01
(MAC and serial available by DM)
Demarc: CommScope SV-4G 4-way splitter, 5-1002 MHz, all 4 outputs populated, ground block present and bonded.

CURRENT READINGS (sampled 12:32 local, 27 Aug 2026)
  Upstream bonded channels ....... 1        (expect 4)
  Upstream channel held .......... ATDMA ch20, 35.6 MHz
  Upstream transmit .............. 47.8 dBmV
  Upstream OFDMA ................. not locked
  Downstream QAM lock ............ 32 of 32
  Downstream QAM power ........... -0.6 to -2.7 dBmV
  Downstream QAM SNR ............. 43.0 dB minimum
  OFDM ch193 722 MHz ............. -4.3 dBmV / 39.0 dB
  OFDM ch194 957 MHz ............. -8.9 dBmV / 34.8 dB, repeatedly drops lock

Note the tilt: -0.8 dBmV at 381 MHz falling to -8.9 dBmV at 957 MHz, about 8 dB
across the band. The 957 MHz OFDM carrier is the only one losing lock.

THE PART I THINK MATTERS MOST: REGISTRATION FAILURE
At 11:41 the modem was restarted. It came back up and could NOT complete
registration. For 21 minutes it held exactly one downstream channel, zero
upstream channels, and no time-of-day, restarting itself twice trying. It only
recovered after I physically re-seated the coax connectors at 12:03.

  11:40   uptime 1:49:40   online, degraded, 32 DS / 1 US
  11:41   restart
  11:45   uptime 0:03:54   1 DS channel, 0 US channels, time not established
  11:50   uptime 0:08:54   unchanged, not registered
  11:55   uptime 0:00:28   self-restarted, not registered
  12:00   uptime 0:00:18   self-restarted again, not registered
  12:05   uptime 0:02:32   registered, after connectors re-seated at 12:03

Events logged in that window, modem clock reading "Time Not Established":

  Warning (5)    B-INIT-RNG Failure - Retries exceeded
  Critical (3)   SYNC Timing Synchronization failure - Failed to acquire
                 QAM/QPSK symbol timing
  Critical (3)   No Ranging Response received - T3 time-out
  Critical (3)   UCD invalid or channel unusable

And earlier at 11:14:
  Critical (3)   Received Response to Broadcast Maintenance Request, But no
                 Unicast Maintenance opportunities received

So restarting the modem is not a fix here, and it is actually risky: the modem
cannot re-register while the upstream is impaired.

FAULT PATTERN
Bursts of 60 to 90 seconds, roughly 110-120 UCD-invalid plus 25-30 T3 timeouts
each. 288 such entries logged so far. Burst start times, local:

  10:18  10:30  10:41  10:53  11:04  11:14  11:19  11:31  12:03  12:14  12:26

Consistently 11 to 12 minutes apart. They resumed immediately after the modem
recovered at 12:03.

IMPORTANT: at 11:05 the modem briefly bonded all four upstream channels, at
39.5 / 39.5 / 38.8 / 47.8 dBmV, then dropped back to one five minutes later. So
the line can support four channels and the modem's transmitter works. The
impairment is intermittent, which means a tech arriving in a quiet window may
see nothing wrong. I have the logged history to cover that.

WHAT I HAVE ALREADY DONE, so it does not need repeating
  - Re-seated all coax connectors at the modem and the wall plate. This raised
    downstream SNR from 41.4 to 43.0 dB and restored registration. It did not
    change the upstream at all.
  - Restarted the modem four times, twice by hand and twice by itself. Does not
    fix it.
  - Inspected interior wiring: one wall plate, one continuous RG-6 run to the
    modem, no interior splitter.
  - Inspected the demarc: the 4-way splitter above, ground block bonded.
    

WHAT I AM ASKING FOR
  1. Sweep the return band, 5-42 MHz, for ingress. This is the main request.
  2. Localize it by metering in sequence: at the tap, at the ground block, at
     the splitter input, then on the modem's leg.
  3. Check the drop from tap to house for shield integrity and connectors.
  4. Inspect and re-terminate the NID connectors as warranted.
  5. Establish which of the four splitter legs actually feed live equipment. Any
     leg ending at an unused outlet is an open ingress path.
  6. Give the modem priority on the split, ideally a dedicated home run, or at
     minimum the low-loss port of an unbalanced splitter rather than sitting
     behind 7 dB of a 4-way.

TWO QUESTIONS I CANNOT ANSWER MYSELF

  - The modem's time-of-day runs 59 minutes 52 seconds behind local time, which
    looks like a stale UTC offset in the config file. Cosmetic, but worth
    correcting.

RESOLUTION CRITERIA
  - Four upstream channels bonded and staying bonded
  - Upstream transmit in the high 30s to mid 40s dBmV rather than 47.8
  - Upstream OFDMA locked
  - Both downstream OFDM carriers locked and holding, including 957 MHz
  - No T3 or UCD-invalid entries across a full hour

On modem replacement: the modem reports 32 of 32 downstream channels locked at
normal power and SNR, and did achieve full 4-channel upstream bonding at 11:05.
I am willing to swap it as a final step if the plant checks out clean and the
fault persists, but it does not address anything measured above.

Happy to provide the full 5-minute log for any window that would help.