I see the though questions arise

1. On DocumentId, I agree with you that range queries are unlikely to be useful. Also, I would like to take advantage of using an ac-hoc FieldType type since we know exactly how we're going to use this field. I think the user should be free to add additional @Field annotations if he needs so, but the default id field which we generate should be custom (and we should keep the luxury of choosing the details of storage options as an implementation detail).

1. I was expecting it to break queries indeed. We can't do much for those who built a Lucene Query using Lucene APIs but we should be able to validate the types when people use the DSL right? For the Lucene Queries, the best we can do is to provide the dump of the indexing metadata to make it easier for people to find out how we're indexing things. Making this dump available ties a bit into the free-form work.

1. Understood. In my mind that's the same problem as in the previous point: let's try to catch & warn about mismatches where we actually can do so.