What about the OP's observation of the failure on the Remote Server?
Is this just wrong, or what needs to be done for the Remote Execution to
be successful, despite the lack of the JAR file, since classes are built into
the JFX file ?
So, specifically, does this mean that a Remote Run can be done
without the need for the @Library designated supporting jars being present
on the server ? ...are ALL classes in @Library references at compilation
time placed in the resulting .jfx ?
IT WORKS
I have to conclude the OP's original observation is just incorrect...
I created a test strategy with a supporting jar referenced by @Library and
verified that for Local Run and for Remote Run, that my custom class file was
indeed loaded in both cases, WITHOUT the need for the JAR file being
present either locally or remotely at runtime; it's only needed at
compile time when the jfx is created.
So the JFX does appear to include at least referenced classes from
supporting jars, at compile time, so that the JAR is not needed at
run time. Not sure about classes or resources which are not
directly referenced by the compiler but will experiment further...
It is clear to me now how you handle the "bundling" of
class files into the JFX at compile time, so that we don't need JAR files
on the Remote Server. Brilliant work !
But, I doubt many realize this, so it would be helpful if some Wiki
entry could be made to really explain the jfx compilation and
subsequent execution environment, with emphasis on local
versus remote execution.
Thanks, and sorry if I am just slow to understand...

HyperScalper