If viewing will be mostly deep sky, then go with 4.
If you plan on using the scope for planetary use, then go curved 3 or 4 vanes.
Why? Is it really better to spread the diffraction across the image rather than have it concentrated in four spikes? I don't think so.
If we are talking about visual use then the sensitivity of the human eye to small contrast reductions across an entire image is crucial.
It seems commonly accepted, for example, that illumination drops of 40% at the edge of field are acceptable since human vision is not very sensitive to reductions of this magnitude. (In fact it is often held that trying 100% illumination across the field is bad design, forcing compromises in other areas in return for an insignificant visual effect.)
Remember that the area of the vanes is small compared to that of the secondary, which is already causing contrast loss across the image.
Can the human eye actually detect the difference between, say, a 20% CO with no spider and a 20% CO + 2% spider vane? Is there evidence that it can?
Asserting that it can seems to conflict with statements in Suiter, and also seems unlikely simply due to the human eye usually being insensitive so small incremental changes of this kind. It is commonly held here that CO obstruction equal to or less than 25% causes minimal degradation. This seem inconsistent with the idea that a 20% CO + 2% spider would cause objectionable effects.
One poster here did an experiment a few years ago with different vane designs (mocked up with masks) and found even very thick curved vanes to be undetectable.
Suiter's analyses show very small contrast reductions from spider vanes.
It would seem that the only reason we would be aware of it at all visually is if, by chance, the entire effect where concentrated in a few small areas of the image and in such a way as to activate the eye in a way it is particularly sensitive to.
And that is what straight vanes do. The diffraction energy shows up in a few tin areas of the image, usually in a high contrast way (black background), and in a straight line for which the eye is a very sensitive detector.
Three and four vane spiders concentrate diffraction and thus render it quite visible, and a potential interfering artifact. It does not follow that spreading it over an image in a way that they eye cannot detect it is as bad or worse.