Contingency handling for data preservation in out-of-space conditions

[fia0]

We might want to investigate how to best handle data loss in a variety of scenarios. This issue mostly stems from the buffering nature of the b-epsilon tree, as more data, than actually semantically present, is stored, with individual copies of data stored at each layer of the tree. A simple fix would be here to congregate data when we try to synchronize the storage state invoked by dropping the database.

Considering one of the major factors of NVRAM, which we want to include soon, is the improved resilience, this might be worth implementing.