One of the many improvements, shipped with the SQL Server 2014, made to the iterators used in the batch mode processing query execution paradigm applicable to a query that uses at least one columnstore index (Apollo) , corresponds to the enhancement and extension of aggregation.

Among other things, it now uses one shared hash table instead of per-thread copy. This provides the benefit of significantly lowering the amount of memory required to persist the hash table but, as you can imagine, the multiple threads depending on that single copy of the hash table must synchronize with each other before, for example, deallocating the hash table. To do so, those threads wait on the HTDELETE (Hash Table DELETE) wait type.

For further details on what the vector-based query execution method known as “batch processing” is, I’d recommend you read Eric N. Hanson’s SQL Server Columnstore Performance Tuning wiki page.