I do:
df.write.format("delta").mode("append").partitionBy("timestamp").option("mergeSchema", "true").save(destination)
If table doesn't exist, it creates new table with "minReaderVersion":3,"minWriterVersion":7.
Yesterday it was creating table with "minReaderVersion":1,"minWriterVersion":2. I changed nothing and now versions are different. Setting delta protocol version on cluster does nothing.
This causes breaking change to me because some tables are used by Azure Data Factory and it's not compatible.