Using the Lightning Data Service - Considerations page as reference, it appears that currently you are able to work with data across spanning relationships (lookup, master-detail) but not collections of records (subquery).
Lightning Data Service supports primitive DML operations—create, read,
update, and delete. It operates on one record at a time, which you
retrieve or modify using the record ID. Lightning Data Service
supports spanned fields with a maximum depth of five levels. Support
for working with collections of records or for querying for a record
by anything other than the record ID isn’t available. If you must
support higher-level operations or multiple operations in one
transaction, use standard @AuraEnabled Apex methods.
After doing some testing, I can say that this is very inefficient. Loading more than about 6 records caused a separate loading to occur. While the system did not crash, as I hypothesized originally in my comment, the total loading time was incredibly large. For about 800 records, it took over 20 seconds to load, about 40 records per second. And this was for a field specified by the fields attribute. This is opposed to my PagingSortingDemo, which loads about 52,770 records with 4 fields in three seconds, or about 17,500 records per second, a total increase in speed of about 43,900%.
The moral of the story is that you should only use Lightning Data Service to load one record at a time. Similarly, if you're going to choose to save data, make a list and save them all at once. Lightning Data Service is ideal for things like Quick Actions that update or query the related record (and with even better performance when using the cache). It is not ideal for bulk loading or saving of data.
Hopefully, they'll introduce a bulk-friendly option for both loading and saving data in the future. For now, you should prefer to use Apex Code whenever you need to upload more than a few records at a time.
Best Answer
As per Salesforce Docs:
SO you need to create multiple instance of it to handle bulk operation which is not recommended.
Considerations