I also sometimes have the same problem with pppm when equilibrating spc/e water systems (although the water model doesn’t really matter, in my opinion) and tweaking the neighbor list and the timestep is not always effective.
Following Axel’s advice, during the first stages of equilibration I turn off the long range interection correction by using the lj/cut/coul/cut pair style. Then I turn it on later, when the system is roughly relaxed. If you have atom overlaps, you could also consider to use the langevin thermostat with a small relaxation time after the energy minimization.
Best,
Davide