[gmx-users] error particles communicated, 2/3 cutoff, domain composition - with a twist

[p buscemi]

Deaer Users,

I've gotten the apparently common error
"particles communicated to PME rank 19 are more than 2/3 times the cut-off
out of the domain decomposition cell of their charge group in dimension x.
This usually means that your system is not well equilibrated."

but with a twist. The 5 ns npt run goes smoothly, the trr is generated and I can view the output. All appears normal. The error is generated while trying to run energy groups with the rerun command. I don't understand how the actual run can complete but the replay does not.
hints would be greatly appreciated.. ~
These are the input commands
;;================================ npt ===============
gmx grompp -f mdp/npt.sys.mdp -c PE.sys.nvt.gro -r PE.sys.nvt.gro -p sys.topol.top -o PE.sys.npt
gmx mdrun -deffnm PE.sys.npt -nb gpu -pme gpu -ntomp 8 -ntmpi 8 -npme 1 -gputasks 11110000 -pin on

;============================= with energy groups =======
gmx grompp -f mdp/npt.sys-eg.mdp -c PE.sys.nvt.gro -r PE.sys.nvt.gro -p sys.topol.top -o PE.sys-eg.npt
gmx mdrun -s PE.sys-eg.npt.tpr -rerun PE.sys.npt.trr

The npt.mdp file for the successful run. the only difference for the energy run is the addition of energy groups/define
================================================npt ==============
; Run parameters
integrator = md ; md = leap-frog integrator
nsteps = 5000000 ; 2 * 50000 = 100 ps
dt = 0.001;

; Output control
nstxout = 50000
nstvout = 50000
nstenergy = 50000
nstlog = 50000

; Bond parameters
continuation = yes ;; no means apply constraints at start Use no if polymer is constrained
constraint_algorithm = lincs ; holonomic constraints
constraints = all-bonds
lincs_iter = 2 ; accuracy of LINCS normally 1
lincs_order = 8 ; also related to accuracy normally4

; Neighborsearching
cutoff-scheme = Verlet
ns_type = grid ; search neighboring grid cells
nstlist = 12 ; 20 fs, largely irrelevant with Verlet scheme
rcoulomb = 1.4 ; short-range electrostatic cutoff (in nm)
rvdw = 1.4 ; short-range van der Waals cutoff (in nm)
verlet-buffer-tolerance = 1.4e-4 ; *****************************

; Electrostatics
coulombtype = PME ; Particle Mesh Ewald for long-range electrostatics
pme_order = 4 ; cubic interpolation
fourierspacing = 0.16 ; grid spacing for FFT

; Temperature coupling is on
tcoupl = no ;V-rescale
tc-grps = HDPE TMPT ;
tau_t = 0.1 .1 ; time constant, in ps
ref_t = 320 320 ; reference temperatur

; Pressure coupling is on
pcoupl = Berendsen
pcoupltype = surface-tension
tau_p = 2.0
ref_p = 1.0 1.0
compressibility = 4.5e-5 0
refcoord_scaling = com

; Periodic boundary conditions
pbc = xyz ; 3-D PBC
; Dispersion correction
DispCorr = EnerPres ; account for cut-off vdW scheme
; Velocity generation
gen_vel = no
gen-temp = 320