Address code review feedback

README.md

@@ -261,9 +261,12 @@ account. Self creation is the only supported way to create an account. If we
 have a new staff member, they first need to login to self create an account
 with the basic role and then admin staff can assign them appropriate roles.
 
-`Thesis Processor` is assigned to any users that process theses.
+`Thesis Processor` is assigned to a user that processes theses. This role is not currently used, however, as our thesis
+processors need a higher level of permissions than it offers. We have retained this role for potential future use (e.g.,
+if we have multiple processors, and some don't need the full set of permissions).
 
-`Transfer Submitter` is assigned to users that transfers thesis files (typically department admins).
+`Transfer Submitter` is assigned to users that transfers thesis files (typically department admins). Stakeholders are
+responsible for assigning this role.
 
 `Thesis Admin` can do everything a `Thesis Processor` can do but can also create
 and update any thesis (not just their own like a `Basic` user).
@@ -302,13 +305,16 @@ Additionally, degrees and departments can be manually seeded from a CSV file if
 documentation for link to a Google doc with the initial list of departments and degrees that were loaded into the
 production database (not maintained).
 
-Seed data is not maintained to match the production database values, which can be changed by admin users as needed. The
-production database *shouldn't* ever need to be reseeded.
+Seed data is not maintained to match the production database values, which can be changed by admin users as needed. *Do
+not ever reseed the production database.*
 
 ### QA/Stakeholder testing data
 
-We're working on a process to load test data for stakeholder testing/QA in an automated fashion. Note this is different
-from fixture data used for automated tests. Check back soon for more info!
+Currently, most stakeholder testing occurs in staging after a feature has been merged, as staging contains test data.
+We're in need of atuomated process to load test data to PR builds for stakeholder testing/QA. (Note this is different
+from fixture data used for automated tests.) This would make it easier to keep the `main` branch deployable while
+multiple features are in QA, and it may help with testing and QA in staging, where records can quickly become invalid
+as the data model changs.
 
 ### Registrar data