GRIP has been around for a while. It does work well enough, certainly easier than OpenCV. However, a Limelight is much more advanced than what you can do just in GRIP. For example GRIP does not do pose estimation, you must do your own. There are several other factors as well that make the limelight a far more consistent and fine-tuned system for most teams.
I have used a GRIP pipeline alongside the WPILib Raspberry PI Image/OS, but it is a much more difficult system to get working well.
Important to note that there were a lot of customizations made to get limelight to the point where it is.
EDIT: An example of something I definitely wouldn’t want to implement by myself with GRIP that is available stock with limelight: http://docs.limelightvision.io/en/latest/crosshair_calibration.html