FREQUENT PERIODIC PACKET LOSS — drop replaced, issue persists — maintenance escalation

TL;DR: Periodic upstream outages every 13–15 minutes — drop replaced, issue persists — need node ingress investigation / maintenance escalation

Houston area customer on a 500mb/s DOCSIS 4.0 plan with an XB10 gateway. I work from home and I'm getting brief, recurring outages severe enough to repeatedly drop my work VPN/remote desktop sessions.

I've done extensive diagnosis and I'm confident this is upstream noise ingress at the node level — I need this escalated to the maintenance/line team, because premises troubleshooting has been exhausted.

The evidence

I ran timestamped pings overnight (~10 hours, ~36,500 pings each) to three targets simultaneously: my gateway (10.0.0.1), the first responding Comcast hop (Greenpoint Houston router), and 8.8.8.8. Results:

• Gateway: 2 lost pings out of 36,533 (0.005% — clean)
• Comcast hop: 770 lost (2.1%)
• 8.8.8.8: 754 lost (2.1%)
• 37 distinct outage episodes, each 2–5 minutes long, with 10–18% packet loss during episodes
• The Comcast hop and 8.8.8.8 fail simultaneously in every episode while the gateway stays clean — so the fault is between my gateway's WAN side and your first router
• Episodes recur every 13–15 minutes, around the clock — 3 AM looks identical to 11 PM. This is not congestion. The strict periodicity looks like cyclic noise ingress from a duty-cycling motor (compressor-type device) somewhere on the node, entering through a shield fault.

What's already been ruled out

• Wi-Fi: reproduced identically on Ethernet
• My equipment: gateway pings stay at 0% loss during every outage; gateway is not rebooting (uptime continuous)
• Signal: downstream power -1 to -2 dBmV, SNR 42–45 dB, upstream TX 36–41 dBmV, zero uncorrectable codewords on all channels
• My own appliances: I manually cycled my AC and compared against episode timestamps — no correlation
• My drop: a tech visited, found a kink in the old drop, and replaced it. The 13–15 minute outage pattern continued after the replacement, confirmed by fresh ping logs.

My interactions so far

Phone support had me restart the gateway (understandable, but it was never going to fix this) and scheduled a standard in-home visit. The tech who came was genuinely willing to help and stayed to troubleshoot, but the visit stalled for a structural reason: he could only show me stats at the modem — the same downstream data I can already see, which is clean. My modem physically cannot see upstream receive SNR or return-path noise; that data only exists at the CMTS or on a meter in the return band at the tap during an episode. Snapshots between episodes will always look fine — the impairment is only present 2–5 minutes out of every ~14.

What I'm asking for

1. Pull the upstream receive SNR/MER and upstream FEC errors for my node at the CMTS — and watch it live for 15 minutes. My outages are on a ~14-minute clock; the next one is never more than 15 minutes away. It will perform on cue.
2. Check my modem's flap list / partial service events for the last several days — 37+ episodes per night should be logged on your side with timestamps matching mine.
3. Escalate to maintenance for a node return-path ingress investigation. The premises side has been fully cleared: clean home network, clean signal, new drop, issue persists.

I have full timestamped ping logs (as well as other non timestamped ones) and an episode table. The 2% loss on my overnight study was actually a decent result. During a given workday I normally see closer to 5-9%. 

A WARNING TO OTHERS : Just use a different ISP. Getting an issue like this resolved with Xfinity is near impossible - you end up wasting your breath on the phone, and with the technician just to get nothing fixed. Im trying fiber from the company that rhymes with PG&E on Monday, and will update this forum with these results.