Thanks to everyone making this possible! Still using it with a very old version, because that's /all I need/ - and it stays the way I like and know it.
Even my 3y old daughter can navigate it using a handheld keyboard/mouse HCI device.
All open standards and compatible interfaces. Even the battery of the HCI is a "nokia BT3" or so.
A long-lasting and ever-stable system. I love it.
Perfectly integrated in my network, all custom the way I like it - even perfectly in-house NTP (LAN) date/time synchronized.
Even custom local-only domain and DNS: all working like a charm for 20 years.
Until one day...
The date and clock sync did not work anymore.
Something had changed, but I hadn't done any change on the system itself for ages.
Years literally.
<Here is where a very long and going several days, weeks, debugging and web-reading and try-and-error: nothing! odyssey commenced.>
I thought I had changed nothing.
Except for one thing: The power-supply.
I've "just before the timesync went arrrrgh! b0rken. ifixit!" - changed the power supply.
To a "better" one. A brand - and brand new - 2A USB power supply because I was seeing the "flash icon" on the RPi more often on the older PSU.
Well, but the "awesome phone charger brand" comes with Quickcharge (Qualcom?) - which /may/ - I haven't spent more time on debugging /that/ - be currently the only reason I can imagine why a proper (huge, btw) USB-PSU from Anker should not suffice to power my RPi3 - who previously ran for years on a 10 EUR random-wahtever-from-the-store USB-charger, even the random Samsung "comes with the phone" charger worked better.
So, how did the PSU mess up the timing?
I saw in the logs (TODO: paste notes from debug-notebook here) that the timesync (connman! btw. not systemd-timesyncd, nor ntpd.
I think there was a "systemctl connman" something available on libreelec.
Anyways: Libreelec's system logs show "alerts about under-voltage" - and therefore slowing down the bus-clock-rate of the RPi hardware itself (so I interpreted the log entries) - leading to a racing condition of connman's NTP part (script where?) and the network interface.
So I propose.
