Simulation crashes frequently when using Dihedral as CV in PLUMED Gromacs

[tongdu...@gmail.com]

Dear PLUMED users,

I am doing some free energy calculation using the dihedral angle, which is defined by center of mass of 4 groups of atoms, as collective variable. But my simulation frequently crashes with the following error information:
-------------------------------------------------------
Program mdrun_plumed, VERSION 4.5.5
Source code file: stat.c, line: 705

Fatal error:
XTC error - maybe you are out of quota?
For more information and tips for troubleshooting, please check the GROMACS
website at http://www.gromacs.org/Documentation/Errors
-------------------------------------------------------
(Both metadynamics and umbrella sampling will encounter this error)

The problem should not be lack of disk space because there is still more than 0.5 TB free disk space in the computer and I don't use quota to limit the disk usage.
But when I check the trajectory before simulation crashes, I find that actually the molecule is dragged into pieces, i.e. the system is exploding before the simulation crashes.
I then try to simulate the same system with the same mdp options using Gromacs without PLUMED, the system never explodes for a sufficient long time.

So my question is:
Is it probably that there are bugs in the dihedral constrain algorithm in PLUMED? And why the exploding system leads to the seemed irrelevant XTC error? And How can I prevent my system from exploding when simulated with dihedral as CV?

Anyone who have encountered this problem? Please help me. Thanks in advance.

Additional is some information of the model system I used to test this error.
I construct a alkane with 8 Carbon atoms, by taking each 2 carbon and its hydrogen atoms as one group, there are 4 group of atoms to define the dihedral used as collective variable.
the plumed input file is as follows:
########
    PRINT W_STRIDE 1000
    TORSION LIST <g1> <g2> <g3> <g4>
    g1->
    1 2 9 10 11 12 13
    g1<-

    g2->
    3 4 14 15 16 17
    g2<-

    g3->
    5 6 18 19 20 21
    g3<-

    g4->
    7 8 22 23 24 25 26
    g4<-
    UMBRELLA CV 1 KAPPA 500 AT    -3.141593
    ENDMETA
########
I am trying to simulate a single molecule in the vaccum. PBC is turned off and there is no cut off.
constrain is set to none, so the time step is reduced to 0.5fs correspondingly.
the related mdp options is as follows:
########
constraints = none
; Neighborsearching
ns_type     = simple        ; search neighboring grid cells
nstlist     = 0     ; 10 fs
rlist       = 0     ; short-range neighborlist cutoff (in nm)
rcoulomb    = 0     ; short-range electrostatic cutoff (in nm)
rvdw        = 0     ; short-range van der Waals cutoff (in nm)
; Electrostatics
coulombtype = cut-off
; Temperature coupling is on
tcoupl      = V-rescale
tc-grps     = System  
tau_t       = 0.1   ; time constant, in ps
ref_t       = 300   ; reference temperature, one for each group, in K
pbc     = no
gen_vel     = yes
gen_seed    = -1
Gen_temp    = 300
#########