The error message
NBONDA>> Maximum group spatial extent (12A) exceeded.
Size is 16.39 Angstroms and starts with atom: 9477
Please check group boundary definitions.
is actually very significant (and should probably be assigned a more serious error level). I means that atoms may be missed when the nonbond list is generated, so at the next update there may be surprises. While increasing ECHECK may HIDE the problemis unlikely to solve it.
Probably, somewhere in the deluge of RTFs that you read there is a residue, which you use in your PSF, with inappropriate GROUP delineations. A GROUP should be small (spatial extent less than 12A per the error message).
It is also a good idea to post a complete AND as brief an input file as possible that exhibits the problem (ie, no unnecessary garbage like using CHARMM variables, doing a number of elaborate manipulations - just find the smallest set of commands and the smallest system that shows the problem)
I strongly suggest to read only the RTF/PARAM files that you actually are going to use; if reading everything all the time were optimal, there would be no great need to have several files ...