Problem with NEB calcualtion

Hello everyone I facing a strange problem while running NEB calculation. I am facing Irrational results. For example I am studying two different defects carbon antisite vacancy(CAV) and silicon antisite vacancy(SAV) formation barrier.
The problem arises with etol for the same geometries with different defect. CAV with etol of 0 give normal result and it’s physical as well while SAV with etol of 0.0 diverges(ebf ebr shoots to 400+ making it unphysical). so TolE never reaches and calculation finishes with tolf. however increases tole to 0.001 allows to finish normally but make it resistive to spring(there is no effect of increasing spring constant) I make sure all the replicas are optimised completely and calculation finished with criteria.
can you go through my input+attachment for SAV and advise me anything.

units metal
dimension 3
boundary p p p
atom_style atomic
variable u uloop 20
atom_modify map array sort 0 0.0
#initial data
read_data sic.initial
pair_style edip/multi
pair_coeff * * /usr/local/Cellar/lammps/20220623/share/lammps/potentials/SiC.edip Si C
#initial minimization to relax whole supercell
timestep 0.0001
displace_atoms all random 0.05 0.05 0.05 123456
minimize 1.0e-50 1.0e-50 100000 100000
#fix 1 all box/relax iso 0.0 vmax 0.0001
#define groups
region surround block -3.0 5.0 INF 10.0 INF INF units box
#group of NEB atoms
#group nebatoms region surround
group nebatoms id 484 1374 1373 1364 1253 557 556 555 548
group nonneb subtract all nebatoms
#apply nudging among replica, fix 1,fix 2,fix3 . specify as many as u want
fix 1 all neb 4 parallel neigh
thermo 100
#save dump
dump 1 nebatoms atom 10 dump.neb.$u
#dump 2 nonneb atom 10 dump.nonneb.$u
write_dump nonneb atom dump.nonneb
#run NEB for 2000 steps or to force tolerance
timestep 0.001
#fix 1 all box/relax iso 0.0 vmax 0.0001
min_style quickmin
neb 0.0 0.001 1000000 1000000 100 final
sic.initial (79.3 KB) (228 Bytes)

As a person that is not familiar with your research domain, I cannot make sense of what you are writing about (too many acronyms and abbreviations) or what exactly your problem is.

If this is an issue of LAMMPS not doing what is documented, you need to provide proper proof for that. If this is an issue of not getting the expected results, then you may be posting to the wrong forum. It is not very likely that there are people here that will review and debug your input, and correct it for you.