Network losing connection erratically

Hi there,

I’ve been running my HomeAssistant as a VM on a ProxMox node for a few years now. Everything has been fine.
Until recently and I cannot put my finger on it when this changed.
I noticed some slow responses to my Zigbee lights connected with the Sonoff ZigbeeBridge (WiFi based). But also some WiFi lights (Yeelight) did not respond on the first request.

After a bit of troubleshooting I noticed that when I ping my home assistant IP, I see an incidental packet loss and a slow response time but mostly it’s running tight at 1ms response:

Reply from 192.168.15.5: bytes=32 time=2429ms TTL=64
Reply from 192.168.15.5: bytes=32 time=2ms TTL=64
Reply from 192.168.15.5: bytes=32 time=1ms TTL=64
Reply from 192.168.15.5: bytes=32 time=1ms TTL=64
Request timed out.
Reply from 192.168.15.5: bytes=32 time=2734ms TTL=64
Reply from 192.168.15.5: bytes=32 time=1ms TTL=64
Reply from 192.168.15.5: bytes=32 time=1ms TTL=64
Reply from 192.168.15.5: bytes=32 time=1ms TTL=64

When I ping cloudflare’s DNS server from the SSH console:

64 bytes from 1.1.1.1: seq=308 ttl=58 time=5.129 ms
64 bytes from 1.1.1.1: seq=309 ttl=58 time=5.112 ms
64 bytes from 1.1.1.1: seq=310 ttl=58 time=4.821 ms
64 bytes from 1.1.1.1: seq=311 ttl=58 time=4.919 ms
64 bytes from 1.1.1.1: seq=312 ttl=58 time=5.044 ms
64 bytes from 1.1.1.1: seq=313 ttl=58 time=5.263 ms
64 bytes from 1.1.1.1: seq=314 ttl=58 time=4.947 ms
64 bytes from 1.1.1.1: seq=315 ttl=58 time=5.117 ms
64 bytes from 1.1.1.1: seq=316 ttl=58 time=5.799 ms
64 bytes from 1.1.1.1: seq=317 ttl=58 time=5.200 ms
64 bytes from 1.1.1.1: seq=318 ttl=58 time=3003.540 ms
64 bytes from 1.1.1.1: seq=319 ttl=58 time=2003.495 ms
64 bytes from 1.1.1.1: seq=320 ttl=58 time=1003.401 ms
64 bytes from 1.1.1.1: seq=321 ttl=58 time=7.472 ms

When I do the same/to from the proxmox host I don’t see these issues so they seem to be isolated in the HASS OS.

I’ve been wanting to move away from the Proxmox method and install Home Assistant natively on my NUC, and maybe this is my cue to do so… But maybe there’s a known issue I haven’t found yet??