Dear Stuart,
The script easythresh_conj is only for Feat, and basically takes computes the parametric & Random Field Theory results to give you the conjunction inference. For FSL-VBM everything is based on nonparametric permutation and thus easythresh_conj won't be of any help, but the principal of conjunction equally applies to any set of tests.
So I have no VBM conjunction script, but the good thing is that, once you have p-value maps, conjunction inference is really easy: Just take the maximum of the two p-values under question. Since randomise writes out 1-P images, you will need to take the minimum of the two images. For example
fslmaths Test1_vox_corrp_tstat -min Test2_vox_corrp_tstat Conj12_vox_corrp_tstat
Since TFCE is voxel-wise-ized cluster inference, you can also do it with that,
fslmaths Test1_tfce_corrp_tstat -min Test2_tfce_corrp_tstat Conj12_vox_corrp_tstat
Cluster-wise inference is a bit trickier; the scope of inference is really on the cluster. Consider if you had two large clusters significant at FWE P < 0.05 and they overlapped by one voxel; it somehow doesn't seem right to call that one overlapping voxel "significant cluster wise".
You can also combine the uncorrected P-values and then feed them into fdr.
Hope this helps!
-Tom