I assume it's related to a bad configuration on host 2 because a 3rd host plugged into 2'socket works fine. This is all wired in the same subnet with a switch in between.
'ping' produces 100% packet loss.
'arp -a' sees systems in the neigborhood with the correct IP and MAC address, but no hostname and sometimes things like:
? (192.168.4.99) at <incomplete> on eth0
route commands produce reasonable results (the same results on host 1 and 2, except for ip addresses(2) vs, hostnames (1)) and host 2 needs about 20s to produce the result vs. host 1 with almost no delay.
The ssh communication works fine - some data packets find their way back to host 1.
There is no firewall involved (rPi-s running raspian)
Host 1: Linux megapi 4.9.80-v7+ #1098 SMP Fri Mar 9 19:11:42 GMT 2018 armv7l GNU/Linux
Host 2: Linux pumpi 4.19.97+ #1294 Thu Jan 30 13:10:54 GMT 2020 armv6l GNU/Linux
Where start debugging? This is a headless system embedded deep inside other hardware. Even swapping the SD card would be rather inconvenient. (poor choice of mine)
$ ip r
default via 192.168.4.1 dev enxb827eb69e3a9 proto dhcp src 192.168.4.50 metric 202<P> 192.168.4.0/24 dev enxb827eb69e3a9 proto dhcp scope link src 192.168.4.50 =metric 202
iptables-save does not produce any output
ip randiptables-saveto your question.