By increasing the interference and there by the hoop stress, you run the risk of coming close to the stress corrosion cracking threshold stress.  

Most engineers I've met use the 50% SCC threshold, that is any sustained tensile stress must not exceed 50% of the SCC threshold.  I'm not so sure about why that number (50%) is so magical, some say it's to acount for additional, non-sustained stresses, others simply because it's always worked in the past.

It's easy to break that 50% rule with high interferences in aluminum and magnesium parts, especially in the ST direction.  I do know of one such failure where the housing hole was in a magnesium part.

Thanks Phil, this helps.

Another reason to keep hoop stress well below the SCC threshold is that the actual hoop stress caused by interference fit is somewhat dependent upon surface finishes, installation methods, etc., and is thus hard to calculate with absolute certainty.  

Also, and probably more importantly, there is often some bending stress due to stiffness variation in the external element; this is usually true in a lug, where bending stress occurs as wall thickness changes.

In addition to SCC concerns, added interference stresses may hasten fatigue failure.

Unless the added stress can be shown to improve the design, as in the case of helpful residual compression, for example, it is usually best practice to minimize the stress.  

Hope this helps.