FENE bond simulation

Dear LAMMPS users:

I used the “chain” tool provided by lammps to create a number of very short chains (3 or 4 bead per chain) for future cross-linking. Although I set the equilibrium length of bonds to be 0.97 when I created the model, there were still a number of bonds which were too long. This brought warnings of “FENE bond too long” but I thought after some energy minimization and MD equilibration all bonds will be enforced to around the equilibrium length. But it turned out that when MD ran for a while and temperature/energy seemed stabled, at some point a few bonds became too long (much larger than 1.5) and the simulation failed. What is the reason for this, bad initial configuration? Do you have any suggestions regarding what is a good scheme to equilibrate a system like mine.

By the way, the potential I used is LJ plus FENE bond with parameters: 30 1.5 1 1

Thanks and looking forward to advices.