One way that may help is first do non-spin polarized calculations and start from this density for your spin-polarized calculations.

Note that ionic (anions specifically) are notoriously hard in convergence. If you are using a Plane Waves basis-set it is very well possible that not all your electrons are located on the atoms in your unit cell (though this is often physical, it is rather annoying), this makes the calculation extremely hard to converge. (eg CH3S- is quite painful, but when you add 2 or 3 H2O molecules near the 'negative S' (simulating the solvent situation) then you will see that the 'left over charge' is allocated on the H2O's and your calculation converges smoothly)...on the other hand if you are using a Gaussian basis-set you may be able to bind the electrons through your basis set (although this would then be an unphysical situation in my book) and the localization of delocalized electrons may be causing problems wrt convergence behavior.

Cationic species should behave better.