I feel Postgres and SQL Server are easier to grok error states. Oracle seems to want to be an application platform more than a datastore
The infamous MSSQL "string or binary data will be truncated" error (fixed a while ago[0]), or the way either deals with unqualified access to columns which leads to surprises (e.g. in MSSQL subqueries or in Postgres's SECURITY DEFINER procedures).
[0] https://docs.microsoft.com/en-us/sql/t-sql/database-console-...
Oracle seems to have more than most, IMO.
I have a journeyman data engineering level experience with each. Rare combo I guess, Data Scientist focus.