First let me congratulate you on being among the "1% elite" of users who actually bothered to find out more about how to use the LISTENER mode in these forums. It was a blast among users back then and still my favorite.
Now on to your issue - I'm assuming when you mention the "sketch I provided" you must be referring to an old code pastebin I've posted way back when I first developed it.
Since then I've made serious improvements in the LISTENER code, in order to ensure stability while running under Spark Pixels control. So I know for a fact it's completely bug-free (running it as we speak, concurrently with many different web clients making and receiving HTTP requests without a glitch). The issue you reported is a very odd one, and couldn't be caused by the LISTENER code. It either would write data to all LED strips or just not at all. Every frame captured over a UDP data stream must be read and parsed; only then its data would be rendered in the LEDs. So, either you've got your matrix mapping kinda screwed up, or those LEDs in that group are failing.
So to help you check I'm posting a few screenshots of how my Jinx is set up for you to compare.
(I've covered the last 2 octets of my local area network IP address) - You can check the cube's current IP address either by loading the Spark Pixels page in your browser and switching from/to LISTENER, then it will show you the current IP address on the top of the page:
(Again, covered the last 2 octets in the screenshot).
Or, if you've got Kevin's Spark Pixels app installed in your phone, you can tap the Main menu then "Particle Cloud Panel" then scroll down to "Particle Variables" and select "debug" then tap "Get Variable".
Output Patch (1:1 pixel coordinates):
(Note the RGB index values highlighted in Red
Output Patch (8:1 pixel coordinates):
Output Patch (8:8 pixel coordinates):
Output Patch (1:8 pixel coordinates):
(Note that the RGB values start from the bottom right towards the top; then continue at the bottom of the left-adjacent column, and so on, until it reaches index 1:1).
If you wanna check the LEDs just try any other mode and watch if the same LEDs are also failing to light up.
As for your other question: true 3D isn't possible because the 3D pixel mapping would have to take place at the source (Jinx) - since Jinx only allows mapping into a X-Y (not Z) coordinate system this limitation is translated into the LISTENER. There are programs which are capable of mapping X-Y-Z coordinates but these aren't free and even the basic license costs a fortune.
I was thinking on making a separate "edition" of Spark Pixels (like a "SP Pack 2" edition or something alike), but user feedback in these forums has been so poor lately I'm not really feeling motivated to keep it up. If maybe I end up making one for myself (privately) I'll share the link with you (PM).
Good luck and post back with your findings.