WDS client: Packet loss with IPv4 but not IPv6?

I have a device configured as an WDS 5GHz client (ZBT-WG3526) to another OpenWrt router (TP-Link Archer C7 v5) in the same room ~3m apart. With IPv4 I have packet loss - only 1%, but with Corona-induced Skype-conferences, that's annoying. Also there is a lot of jitter.

The strange thing is: with IPv6 over WiFi that's not the case. Also IPv4 over LAN works fine:

WiFi IPv4

100 packets transmitted, 99 packets received, 1% packet loss
round-trip min/avg/max = 0.941/51.185/410.563 ms

But with IPv6

100 packets transmitted, 100 packets received, 0% packet loss
round-trip min/avg/max = 0.401/0.484/0.668 ms

Ethernet IPv4

100 packets transmitted, 100 packets received, 0% packet loss
round-trip min/avg/max = 0.435/0.528/1.039 ms

I'm new to this whole wireless business (just being reminded why I love my cables) and have no clue what's going on here. WiFi seems stable, because IPv6 works fine. IPv4 seems stable, because Ethernet works fine. Any ideas of what's going on here?

The WDS client is my LTE router, sitting on the window-sill and I would really hate to have an Ethernet cable dangling across the room...

Thanks in advance!

Which host is pinging which host? The WG3526 to the C7? Vice versa? Some other host?

The client (WG-3526) to the AP (C7). But the effects are noticeable when my desktop connects with IPv4 through the C7 to the WG-3527 and then to the Internet as well.

I would run tcpdump with -tttt option on both devices and verify from the timestamps where is the delay.
If for example the C7 gets the ping and sends the pong much later, something is wrong there. If it sends it immediately but WG3526 gets it much later then the delay could be on the physical layer. Or if WG3526 gets the pong in tcpdump but doesn't display the reply or the reply is delayed, then something is wrong with it.
For example:

2020-07-04 21:42:50.422972 IP odroid.mrv > magiatiko.mrv: ICMP echo request, id 6379, seq 1, length 64
2020-07-04 21:42:50.423074 IP magiatiko.mrv > odroid.mrv: ICMP echo reply, id 6379, seq 1, length 64

Router responded 102μs after receiving the ping.

Thanks, I'll look into that next week.

But it's still kinda strange. The physical layer seem to be just fine, as proven by IPv6. And display also does not seem to be an issue, as there are "real life" losses with e.g. Skype (trainees complaining about second-long dropouts).

Could it be related to WDS? My C7 is a "normal" (non WDS) WiFi client to a provider-supplied Router without any problems, even though it's on the very crowded 2GHz channel 3?

Let's wait for some test results. I cannot imagine what could be the problem by the looks of it.

in my experience WDS would just bork randomly. I had a set up with two APs doing wds from a central wired AP... this would work fine until it didn't... and there was no obvious rhyme or reason to it.

I moved to adding a wire for one AP and a power line adapter to wire in the other location.

I'd look at powerline for your scenario, assuming the LTE router is plugged in, it must have power nearby. run the Ethernet cord to the powerline... it's not dangling across the room. then switches and soforth to give you lots of wires.

in terms of real-time conferences wires are the way to go