Spike across all BOLD summary statistics in fMRIPrep at beginning of scan

I’m preprocessing task-based fMRI data using fMRIPrep, and while the pipeline completes without errors, I’ve noticed a consistent issue across the majority of runs.

In nearly every subject/session, there is a spike at the begining of each run in all major confound metrics (including FD, DVARS, and GSWM). This spike appears isolated but significant. My questions are:

  1. What causes this?
  2. What will it affect down the line to have this inital spike?
  3. How can I clean the data either pre or post fMRIprep to ensure that this doesnt affect the results further down the pipeline?

Thank you in advance