Ongoing DOCSIS 3.1 instability – correlation between downstream power rise, error rates, and latency

I’m looking for a second opinion from those familiar with DOCSIS 3.1 behaviour, as I’m seeing a repeatable pattern that doesn’t look like a simple in-home issue.

Context

• Virgin Media HFC (Hub 5, DOCSIS 3.1)
• Long-running intermittent latency and packet loss
• History of T2/T3 timeouts, upstream profile changes, and partial service events
• Multiple engineer visits; most recent visit (26/11/2025) resulted in an equaliser being fitted due to limited downstream headroom

What I’m seeing
Over the last 12 hours (03/01/2026), the following appears clearly correlated:

• As downstream power levels rise,
• Pre-RS error rates increase significantly, and
• Latency and packet loss increase on the ThinkBroadband BQM

This is visible when comparing:

Grafana modem stats (downstream power + Pre-RS errors)

 BQM (latency, jitter, packet loss)

The rise in errors and latency tracks the downstream power increase very closely. When downstream power stabilises or drops slightly, error rates and latency reduce accordingly.

Additional observations

• Downstream SNR remains broadly stable
• Post-RS error rates remain very low (errors are being corrected, but at scale)
• Upstream power is already high and leaves little headroom
• No corresponding increase in local traffic or load

Behaviour is time-dependent and not tied to in-home activity

Why I’m concerned
This doesn’t look like:

• Home network congestion 
• Local wiring or CPE fault
• Random noise bursts

Instead, it looks more like:

• A plant-level or cabinet-level issue
• Possibly thermal drift, attenuation imbalance, or noise ingress affecting DOCSIS margins
• A condition that won’t reliably present during short engineer visits unless the timing is right

Question
Am I right in thinking that:

• Rising downstream power pushing channels closer to the top of acceptable range could be reducing error margin, leading to increased FEC activity and latency?
• This points more towards a network-side balancing or plant issue rather than anything further an in-home visit can resolve?

Grateful for any technical insight or confirmation from others who’ve seen similar DOCSIS 3.1 behaviour.