@DR: If the variable is a real & it is not initialized then yes it can pose problems, I have seen it, the PLC will stop on error (well sometimes), it shows it as NAN (Not a number) however, initially all the registers seem to be zero when first use after downloading a program & parameters, the compiler will not complain as the variable data is not included in the compile as such, just that it is a variable that resides in the PLC, you have to remember that in Mitsubishi all variables reside as set areas in the PLC memory map so are not generated at compile time with the exception of those used purely as symbolic only to be compatible with IEC format, these in the past cannot be addressed from the HMI as it would not know the address, however, I do believe there is a move towards being able to download the symbolic table so that symbolic addressing in the symbols table contain the actual physical address, Mitsubishi at this time only has fixed tables of memory unlike RSL which was a complete change from the older platforms so being able to directly address symbolic rather than addresses. I believe if Mitsubishi do go the whole hog then this will be a completely new platform.
I have been using Mitsubishi for over 35 years & only seen "NAN" 3 times, once because a D register declared as a real (2 D registers) where the 2nd register was mistakenly used as an integer so it effectively corrupted the real value.
The other times I was unable to find what caused the errors, a simple force to 0.0 or some other real number solved the problem that never occurred again.
In general, this would only happen at first download of program then putting the PLC into run.
Mitsubishi have for some reason try to keep their IDE's compatible across the older & new PLC range, does causes small problem but rather that than having to pay for a number of different IDE's for the different platforms.