Yup, so take the 4-8GB backup tar file, clean install the new device, restore the 4-8GB backup from the tar file. You'll be done 12-18 hours faster than block cloning a 128GB SD card. Users constantly obsess about cloning, but you don't need an identical card, you only need one with the same data on it, and that's the easiest/fastest route to that objective.
NB: As you discovered, most cards do not have exactly 100% the same 'disk' geometry which complicates block cloning.
Use connmanctl to review and perhaps change the interface sequence order to modify routing.
While 98% of the boxes out there have the same Broadcom RTL8211F chip inside, there are occasionally other chips used, and they can be using a different bus slot which requires a different device-tree and/or maybe an additional driver in the image. As a start, share the system log (assuming you can access SSH over WiFi).
The wiki is open to public contributions.
The solution is to back the contents of the card and not clone (at block level) the entire card. This will also be massively quicker as you're usually backing up no more than 4-8GB of Kodi data, unless there are media files on the card, rather than the entire 128GB card. Learn to use tar.
I've pushed an updated set of images to my test share with:
I'm interested in any reports of HDMI sync loss when playing media (along with mediainfo details) as the fix for YUV420 @ 59.94 fiddles with HDMI clock plumbing.
I'm also interested to hear if WeTek Hub boards are booting fine now? - some folks had an issue we suspected of being noise on the UART that caused it to see a keypress and enter the u-boot console instead of booting. Feedback would be nice.
WireGuard requires the host (server) to be an IP address and ConnMan does not implement logic to resolve an FQDN to an IP before configuring the WireGuard interface and bringing it up, or perform periodic validation of an FQDN to detect IP changes.
It's not an impossible problem to resolve [sic]. Create a script that resolves the FQDN against a trusted upstream DNS server (not a local one which caches results) and compares the IP to the current WireGuard conf IP, and if different, sed the value in the conf and then restart the WireGuard service to recreate the connection with new values. Use cron or a systemd timer (which is cron) to run the script at periodic intervals.
I made an effort about 2-years back to clean-up the Linux 3.14 era drivers so they compiled against a modern kernel. These can be found here: https://github.com/chewitt/linux/commits/dvb-sucks and I rebased them against Linux 6.12.y only last week. However there's no Amlogic demux driver in the upstream kernel (which is essential) and we need tuner/demod drivers to be independent rather than interdependent so we can describe a box configuration in device-tree rather than compiling everything into a single monolithic driver as the Amlogic vendor kernel does. The monolithic approach works for box manufacturers who need to support a single box with a single driver recipe, but doesn't work or scale when you need to support a range of devices; each with a different tuner/demod/demux driver combo.
It's not a huge amount of work to get done, but it takes a certain skillset and mindset to hande the code archaeology and writing of modern kernel drivers. It's already been 'years' and nobody volunteered, so I have low expectations of it ever happening.
The 12.0.1 release notes that nobody reads describes some "no" items for the AMLGX image:
The latest and last release that might support them is LE 9.0.1 although dtech has an LE 9.2.8 image that might also contain them (but I don't track it, so no guarantees). The upstream kernel in LE 12/13 has reasonable support for USB tuners.
You can start with a Kodi debug log, but I suspect that's not going to be enlightening. If you're able to test and replicate the issue with Kodi running on a general purpose Linux OS like Ubuntu that would be helpful to eliminate LE itself from suspicion and allow you to concentrate on the add-on; which will have a support thread in the Kodi forum where the add-on authors hang out.
From the device-tree name, it's probably the one I created some time ago (used uncredited, as is normal for ophub). See comments in the other thread you created.
MatteN is probably on the right track with his answer. Remove EDID captures and connect the RPi4 board directly to the TV until you get things working. You probably need to experiment with different TV ports and/or port settings to ensure it can accept 12-bit 4:2:2 input, which is what the RPi board uses when 10-bit YUV 4:2:0 (which it cannot output) is needed by media.
See https://wiki.libreelec.tv/configuration/4k-hdr#id-4k-60hz
Note that RPi4 (not RPi5) also requires 4K modes to be enabled in config.txt. You do not want to force the Kodi desktop to 4K60 unless you want the UI to be slow and annoying. Mode switching for playback is fine, but the UI is best in 1080p.
I'd guess that "no signal" refers to the HDMI input signal and when the LE device is powered on the HDMI input is being switched, but the TV fails to lock on the resolution and you see an appropriate message. Killing the antenna signal sounds really weird and I heard nothing like that before.
No idea what the issue is (although Synology and SMB are 100% nothing to do with it) but, a) the better place to ask the question is the Kodi forum where the developers who maintain the Music side of Kodi and the scrapers hang-out, and b) for any commercial media that you've ripped, https://picard.musicbrainz.org is the fastest route to correctly tagged media that scrapes without issue.
The FD628/TM1628 driver was submitted upstream and then bike-shed'ed to a standstill, so it remains un-merged and the author has no plans to resurrect the effort. I have been including patches in the Amlogic kernel that I maintain for the AMLGX image for some time. To have the driver work, the VFD layout needs to be correctly described in device-tree. I've created device-tree files for other boxes in the past, but the only one I still have in the AMLGX image is for the TX3-mini. The driver is not designed for use with openvfd or lcdproc userspace tools, although the openvfd repo is useful for figuring out the pins that need to be bit-banged to make things work. The nearest you can get with userspace configuration is issuing a sequence of fdtput commands to modify the device-tree file that boots the box (which updates will overwrite).
This kernel branch has the commits to add the driver, and a device-tree for Tanix TX9-pro with the VFD described:
Note that this was 10x kernel releases ago and I've no idea now if these patches were ever proven on a real TX9-pro, or whether your TX9 (non-pro) clone will need/use the same layout; although I'd guess the spec difference is limited to RAM size and WiFi/BT module used. Typically users show-up excited that AMLGX exists and then go quiet once they realise playback is not as polished as the older vendor kernel images (not that these are perfect either) so not all device-tree files that I create get tested and sent upstream.
The device-tree compatible also needs to be in to https://github.com/LibreELEC/Libr…s/vfd-clock#L25 .. although it looks like I've done that for the tx9-pro in the past and never cleaned it up.
There are build instructions for creating your own LE images in the wiki if you're that way inclined and want to experiement. I can also pick the patches back into my LE13 test images; although I'm currently working on a kernel driver fix for a long-running bug and it might be a few days before I have that resolved and release a public image again.
I'd guess that if you have Kodi configured to play the next episode and media is opened from a view that lists multiple episodes .. it will do what it's been told to do.
No ideas on the stop/exit although Kodi should buffer if the connectivity is poor, so it sounds more like an issue with media itself or something server-side or how the add-on works 
Random guess based on the PCM comment: have you put Kodi settings from Standard > Advanced or Expert to expose the audio formats to be passed-through?. If none are selected you'll still get PCM.
RPi5 2GB is more than enough for LE unless you want to run a ton of persistent add-ons and services like Docker containers.
