fMRIprep to CONN: data from sites with different resolution and TR

I am working with a resting state data, preprocessed in fMRIprep. The data is from multiple sites and the resolution of functional data varies between them (voxel sizes: 2.5x2.5x2.5 or 2.4x2.4x2.4), as well as TR that ranges from 800ms to 1450ms.

The preprocessed anatomical data is in MNI152 space (1mm). Functional data in native space after resampling on the MNI space.

I want to use the CONN toolbox for ROI-to-ROI analysis, and some as between-groups comparisons with a custom ROI template that corresponds to the fMRIprep MNI152Lin6Asym space, would you advise resampling the ROI template to different sizes to match the native space of the fMRI data?

Or use “:res-2” output option and the bold preprocessed files with unified resolution in CONN?

Are there any red-flags I should be careful with if using any of the above solutions and are there other ways to deal with the discrepencies in TR and resolution?
Thank you,