Ian Jackson writes:

Equally bad on ethernet I expect? Also, does ethernet require them to be
non-directional (ie just resistive tap-offs, with no directional
coupler)?

Ethernet required a direct connection to the conductor of the cable, if
I remember correctly. The receiver side was high impedance, so it
didn't present a significant load to the signal, and the vampire tap was
designed to create only a small hole through the shield and inner
dielectric so it wouldn't produce much of an impedance bump. Plus the
cable was marked with rings to indicate where you could put a tap
without having multiple taps end up a multiple of a wavelength apart.

When transmitting, an Ethernet transceiver acted as a current source,
putting a fair bit of current into the 25 ohm load (50 ohm cable heading
off in each of 2 directions). If two transceivers decided to transmit
at the same time, the high DC level on the cable was used to detect a
collision.

Original Ethernet used a baseband signal, and on a moderate-sized
network every station listened to every other one directly. There's no
"head end" to echo upstream signals back downstream again. There's no
notion of "upstream" and "downstream".

Dave