Thanks for the reply.
This appears to be a known problem, discussed a few times in this forum e.g. Wifi and Bluetooth Have Never Worked - #35 by choboDOC and further link from there.
I don’t know the details of this problem but I wonder whether the only full workaround is to upgrade the WiFi card from Redpine to SparkLAN, which may or may not be something that you would contemplate doing.
I’m wondering, if toggling power fixes it and it’s a firmware issue, if the phone can’t automatically repower the Wi-Fi chip until it works at boot.
I indeed notice more recent batches have the SparkLAN chip, but it’s a company phone so replacing it would need some ACKs. I also can’t imagine it can’t be worked around since Linux supports so many different Wi-Fi chips with their quirks. I also saw a note in another thread that Redpine was planning to pick up mainline Linux support again a few years ago, so that brings some hope I guess.
I would probably use
ping -s NNN ...with progressively higher packet sizes to see whether the problem is not the type of traffic but instead the packet size.
Good idea, I will try that next time, thanks.
It may help to clarify though what you are pinging and which computer is doing the ping.
Indeed, I see how this was confusing. The above was with my laptop doing the ping whilst connected to the hotspot of the Librem which was tethering that to 4G.
What about tether via WiFi? I haven’t done a lot of tethering but I think that’s how I have done it.
Yes, that does work, when it works, i.e. with the aforementioned issues of the AP and Wi-Fi on the Librem dying every so often 
.
When I can get the ethernet adapter to show up and the wired connection to be recognised (which is now almost never due to the above issues), wired tethering over USB-C is noticeably more performant for me though. It works out of the box through the GUI and I originally started investigating it because Wi-Fi was so unreliable, because I remembered how reliable the cable connection used to be .