There are MR-compatible (carbon based) fiducials out there now that are small. But I have always used 1x3mm gold seed markers, and I can get them to show up on T2 turbo spin sequences (I get a rads nerd to help me set the particulars) just fine. They do not distort the MR images. There is almost no artifact. Wait at least a week after marker placement for the MRI and there will be no bleeding issues. I googled DELINEATE and it looks like they put in (3... I have always used 4...) gold markers prior to MRI?
In Eclipse (and I'm sure other TPSs), you can put in matching points and do image fusion that way (versus image-based). The seeds become the match points. This method gives me high confidence in MR/CT match w/ fiducials and would show near-zero inter-user MR/CT match variability.
The argument for MR-fusion for all prostates is that I feel it makes some change to my prostate contouring in at least about 1 in 3 cases. Amar Kishan showed us (I already knew it, natch!) that even a couple mm of contouring difference can affect changes in toxicity outcomes.