judging from g.fast generations performance and alikes - i think copper still serve LOOONG time, atleast for last 150m-400m-1km "last mile" :=)
same bout DOCSIS 3.1 and other legacy things :=)
ana R&D and improvements and standardization - keep flowing, albeit slower
anybody thinking this way about copper pair surely has not see the problems about isolation on outdoor cooper wires or/and induction ac or dc problems
i think docsis has a better future because is more resistant to interference
not over optical networks, especially conventional "active optical networks", reasonaly dominating market for number of advantages.
as for g.Fast and g.Fast 2 comparisons - i think at comparable speeds - amount of interference would be minimal unless you're talking about Very huge/dense cables and/or very high speed, yet unusual even for g.fast endpoints, sadly.
also ethernet had drawbacks over docsis only if you use unshielded cables, which is allowed only for OFFICE used Only(NOT for residental deployment. NOT for outdoor deployment. NOT for "last mile" cable/conversion).
and STP and Cat7 cables - quite EMI-resistant and emit - Much less. same about 6e derrivativers with improvements.
so far DOCSIS is extremely handy as "legacy conversion" thing where problematic technically to lay any more cables at cost of Very segmented networks and reduced # of consumer per aggregating modem/gate, its suffer anyway with HUGE issues both on L1 and L2 by basic nature of this technology, implied by both shared medium and control methods. however in DOCSIS 3.0 and 3.1 derrivatives - they shifted away from CSMA remnats compltely(TDMA and S-CDMA mix in upstream - considered relatimely low-latency with predictible response)to slightly reduce issues on L2 and frequency management in DOCSIS - improved in 3.0 aswell as BPi/SEC improvements, jointly developed with alikes in WiMax - contributed(along with other ethernet sig's memberss)to all 802.1x development/improvements and lead to 802.1x-2010 creation eventually(and all those "PortSec", "MacSec" things and other stuff aside that, ajaced around 802.1x-2010 which ough to reinfroce it Further).
if you had to "put DOSIS 3.0 to extremes" in terms of speed - amount of customers per aggregating modem - become quite small, up to 16 or even 8 per one and todays 64(todays even 512 isn't uncommon in some networks)become "not a option" anymore to achieve "advertised performance" and to go Further "modem per block" or even "modem per building storey" become a MUST. DOCSIS did not scale much and going up only by bonding and bonding imply pre-occupied/converged amounts of HW/spectrum/time "per customer" become bigger in modem-space/freq/time.
from angle/point you speaking i think technology like HomePlug would fit well. but modern implementations like ITU-T G.hn developed VERY slowly in favor of DOCSIS and alikes. which isn't good much
HomePNA 3.1 also support PHY's working over Coax, aside twisted pairs(random of. from phone lines, power lines and whatever else mix of 2-8-12 pairs(depend devices used for)).
so G.Fast-grade HomePNA -derrivative for Coax would be better, yep. if you insist on such medium for some reasons and dislike Ethernet derrrivatives like HomePlug/EoC.