Hi Dustin,
This definitely helps with understanding the nature of the problem better. I have seen this before. I believe TBC's processing engine, after computing a "survey quality" solution for various points (e.g. PID 14), it runs the baselines back out to the each of the static base/CORS stations and finds that they conflict. It makes sense to do this in a way, because it helps you identify conflicting observations for overdetermined positions without running any adjustments, but it can also be frustrating and misleading. I run into this a lot when mixing data types when Network RTK is involved, too.
I have 2 additional considerations:
- If you go to Project Settings > Computations, what is your current setting for "Sideshot computations"?

- Have you considered running a Network Adjustment?