i touched on this in a longer comment in this thread because i think that docsis 3.0 goes up to 900-1002mhz

if downgrading to docsis 3.0 (or downgrading to 500/700mb) “fixed” your issue, you probably have a 5-1000mhz splitter thats not just a rf splitter but also, a filter and its JUST leaky enough to allow 1002mhz through.

or maybe the modems happy negotiating down to 900mhz.

but maybe not quite enough for 1008-1100+ required by docsis 3.1

there will be anecdotal reports of a 5-1000mhz splitter “working just fine” maybe that ones a REALLY leaky filter thats also allowing 1008mhz.

or also a case of negotiating a lower channel…

gigabit speed and docsis 3.0 are about the threshold for the 5-1000mhz

problems would manifest with docsis 3.1, gigabit speed(maybe) and then almost guaranteed at 1.2 gig service+

this idea of “sensitive channels” is extremely close to nailing it

splitters fail as well. they’ll bleed through AND filter bands theyre not supposed to. but i didnt seize on that or inside wiring for OP because “the neighbor gets it too”

im on a gigabit implementation that has to have +/- 1100mhz , and my own woes uncovered an 800mhz splitter inside a wallplate. it would lock. it would even run at gig somehow. just not very well. its a 5-2500mhz splitter now. a 5-1200mhz would also do (for now)

everyone on your tap should be using multiplexed signals, and you should have a good 300mhz or so to play with and lock onto. but if every single one of you gets kneecapped at +/- 1000mhz, then theres a really congested 100mhz band and another 100-200mhz thats open for everyone but you cant lock on to it.

In my case, one of the main loops had a cable break under a roadway, in a place on the loop that was wholly Comcast infra rather than subscriber. You’re not wrong about the general case! But for that reason, they basically stopped acknowledging the issue to me at all, never followed up on support calls ever again, and it took them maybe three years to close that roadway overnight and fix their cable. (I was able to manifest the issue at a service speed of 125mbps when capacity up to 1+ gbps was available, but of course that low limit didn’t stop the modems from negotiating whatever full-width max-QAM links they could.)

Diagnostics mastery note: logically ruling out a readily testable possibility is only (somewhat) logical when one hasn’t exhausted all other possibilities. Displeasing and successful diagnostic tests that ought not to differentiate but do are how one exposes issues hiding in the blind spots of other experts. (If they hadn’t explicitly said ‘I have no ideas left’ in as many words, I probably wouldn’t have posted at all.) Here is an idea they hadn’t openly said they considered. The reasons this idea might or might not pan out are still interesting to me! TIL! But it was a beautiful and consumer-accessible scalpel of diagnostic and earned me a walkthrough of the signal contamination specifics by the senior truck tech who showed up to help the lesser truck tech, so perhaps it’ll help another.