I have visually assessed a number of networks configured by WGCNA. The one that I am most happy with however only has a scale-free topology fit of 0.6 R^2.

As follows:

My downstream analysis of this is to correlate external trait-info to this network: how sure can I be of any inferred correlations if my network is of such a low R^2 fit?

The fit of R^2 is not indicative of how trustworthy correlations of module eigengenes with external traits will be. Regarding low R^2, point 6 (I can't get a good scale-free topology index no matter how high I set the soft-thresholding power.) of WGCNA FAQ has some discussion.

If you have made sure that there are no technical or uninteresting biological drivers of global differences, you have used a reasonable soft thresholding power, and the modules look like distinct branches on the dendrogram, the modules are likely pretty stable and correlations with traits are trustworthy, inasmuch as you trust your data.