MoCA and MoCA-compatible splitters and drastic reduction in bandwidth

I am (for no good reason) renting the Cisco DPC3941T.

 

My network topology is:

 

cable from street -->

  MoCA PoE filter -->

    split +--> split +--> gateway (is MoCA hub, i.e. MoCA is "on")

          |          +--> TiVo Bolt (is a MoCA client, can be hub)

          +--> TiVo Mini (is a MoCA client, can't be MoCA hub)

 

That's two splitters.  Each is currently 5-1000 gHz and was supplied by Comcast back in the day.  3.5 db of loss per out.  I get a massively strong signal and great bandwidth.

 

MoCA wise, this mostly works fine, despite not being MoCA compatible (a MoCA compatible splitter goes "up" to 1675 gHz if I remember right).  But whatever, sometimes this works fine.

 

Then, periodically for no good reason, sometimes weeks at a time, this router decides that MoCA is disabled.  That puts both my TiVos off the network.  I can select "enable MoCA" and if it's the right day of the week or the barometer is looking good it might work.  Most of the time it does not.  Reboots are unpredictable.

 

It occurred to me that perhaps this is because my splitters are not MoCA-compatible and perhaps the router is selecting some sort of MoCA frequency that is being blocked or almost blocked.  So I ordered new splitters that are MoCA compatible.  (My understanding is that MoCA 1.1 and higher devices can "see" if that higher band is available and will use it if so.)

 

When I replace the ordinary Comcast-supplied 5-1000 gHZ splitters with the MoCA compatible ones, my effective bandwidth falls off a cliff and the signal to noise ratio on my downstream connection plummets.

 

By "effective bandwidth", I mean the real-world measure of using the Google Speed Test (search for "speed test" and click the blue "Speed Test" button that is heavily promoted) on my WiFi 5gHZ connection, I go from about 175 to 30.  Eek.

 

For the time being, I put the Comcast splitters back and everything is hunky dory again.

 

What theory can account for this?