I  Think controls should be more than  cases 

Need at least as many controls to justify claim of association for rare-variants relevant for complex phenotypes -- any such background allele frequency estimates require adequate sample sizes

1:1 sampling makes sense if cases and controls are equally difficult to find. In general, for rare diseases the matched controls are easier to recruit. Some power is gained in have up to 4 controls per case.

At least 1:1 sampling is must if controls are not kept high