The purpose of checkpoints is to recover after a failure.  In your case, the streaming query is changed.  Structured streaming isn't stateless.  This means that in general the checkpoints cannot be reused when the query is changed.
Depending on the case, you might be able to recover semantics after changes:
https://spark.apache.org/docs/latest/structured-streaming-programming-guide.html#recovery-semantics-....
In your case the schema changes:
Structured Streaming automatically checkpoints the state data to fault-tolerant storage (for example, HDFS, AWS S3, Azure Blob storage) and restores it after restart. However, this assumes that the schema of the state data remains same across restarts.