I don't really have any good answers for you, but feel your pain. We also have run into this issue with pipelines.
Also we have old notes that are now filled with diamond question marks where a curvy apostrophe or quote, or a micron symbol once was. Sometimes I'm not quite sure what the character originally was. We also have thousands of these so it's not possible to fix them all. I was able to fix issues in a couple of tables that only had a few instances, but not in all the tables.
One thing I would check is to see if you have any of these characters in the field names of the tables, because I think removing them from field names can make a difference even if the data itself holds these "special" characters.
This now just popped up for me where the Pipeline is failing due to a single character in the csv load. I tried for force UT-F encoding but same issue. Is there a setting that will force load the file without rejecting it for encoding issues?
We subscribe to other big platforms, and Quickbase is the only one that continually has these issues. I spend more time working around clunky UI/UX stuff than actually building apps these days. It's too expensive of a platform to make their users do this with little to no support.
