The C++ code of XED access the IQ, RQ, and LQ arrays to be able to access things like the PMT banks, to get calibrated PMT data, etc. This means that it needs to know all the relevant bank offsets: data word definitions, link structure, etc. C++ code, of course, cannot use the bank_XXX.inc files used for Fortran, so all of these constants (e.g. KPMT_PN = 3 ) are stored in a C header file called snoman_mnemomics.h. If any of these values are changed, this file must be changed as well. (Hopefully, this will not happen too often!)
At no point does XED modify the SNOMAN data structure in any way; it treats these arrays as read-only, so no harm can be done even if these mnemonics are incorrect.