Is there a conventional guideline to upsample rs-data?

Is there a guideline to sample rs-data to a different resolution for analysis - if so, why?
I am dealing with 7T rsfmri data and we have a native resolution of 1.2mm. For analysis we have upsampled to 0.5mm MNI space and I am wondering if this will introduce too much noise or cause issues with conventional statistical software assumptions. I know that sometimes 3T data is also upsampled, but maybe not as much? Any input would be appreciated!