Hi @Kirankumarbs 

Please do reconfirm if my understanding is correct 🙂
Summarizing above points:
So checkpoint stores target schema (schemaLocation) and source offset that is delta table version in this case(offsets folder). So on next microbatch run it starts reading from n+1... current version of source if any version includes changes that are non additive it stops and throws error and if it goes well, at the time of write it compares schema of df with target schema.