That +2 was an early attempt to pad things and the end result doesn't improve. It is now removed and the original post updated as well.
This is the simple example code. The change included making it simpler and instead of sending a float, sending a uint8_t and hard coding radio and TX vs. RX for the device I was flashing to avoid typing each time in the command window.
I made the payload size 32 bytes as well and still sent the uint8_t with same results.
I am not sure how much simpler I could start. This code works fine if the architectures on both ends are the same, so I am trying to make this as easy as possible and appreciate the help to troubleshoot.