I created a ticket 2 weeks ago. An ESRI rep got hold of me and I'm not impressed with the answers- Take the input feature and edit in field data design and all outputs will contain / retain the correct field schema.
That answer defeats one of the main benefits of Modelbuilder, the model processes take care of all manual processes. ESRI's answer implies that Parameter of the model has to be fiddled with every time a new dataset is used.
A large chunk of my models deal with taking county centerline and address points, and crossing over to be used in Emergency Service softwares. Some sites I grab new data monthly, point the model at the new data (data fields don't change) and let it rip - historically, works every time!
The 1st part of all models export the agency data into a new feature class where unnecessary fields are dropped and appropriate coordinate systems are assigned. Most models have many processes and uses either export features, merge, spatial join, buffers, feature to point - all of which produce new feature class outputs where there is additional field edits/ modifications being accomplished within the model. PRO 3.2.0, 3.2.1 just blows all field edits / modifications out the window in every new feature class. ESRI attempted to contact me to experiment but after the original example I immediately reverted back to PRO 3.1.4 and will stay there for the time being.
To state it again- None of us should have to fiddle with field design with a new dataset in a repetitive workflow model process. I'm really upset over this.