Developer here.
LightBurn internally works within tolerances that were chosen because they well exceed those of the hardware that outputs the designs, and it results in both performance and memory gains which were deliberate. A lot of users run old hardware, like 10 year old laptops, that are both performance and memory constrained, so using double precision floating point for everything would incur a significant hit on those systems. Given that the hardware we target has an internal precision limit of about 10 microns, keeping our precision tolerances at or lower than that seemed sufficient. I can’t imagine a burned edge piece of plywood holding 1 micron tolerances - Do you disagree?
LightBurn was never designed to be a CAD package - It’s principal reason for existing was to take files from other sources, like CAD or illustration software, and prep it for use on a laser. Over time, significantly more functionality has been added, but it still has limitations, and many of those are intentional.
One issue that was recently fixed is the conversion of positions from internal float to disk format, being done at six decimals of precision instead of eight. (Eight is the limit of the internal floating point representation, six was the default of the “output a number” function)
When you duplicate a shape, it’s the internal in-memory version of the shape that is copied, directly, and dropped exactly where the original was. When you Copy/Paste, it goes through the serialization routines to put a copy on the clipboard in text form, and de-serializes back from that, and so was subject to the 6 instead of 8 places of precision bug.
The beta John attempted to point you at includes this fix and a number of others, but you dismissed it out of hand. That version is really just a bunch of collected bug fixes, and is due for release today or tomorrow.