How does Swarm handle interrupted transfers (abandoned uploads)?

For a file upload, a Swarm node returns a UUID only after it successfully receives the final byte; without it, no object is saved. Regardless of success, the node had preallocated the intended size of the upload and set aside the disk space to store the object.

Through normal health processing, Swarm will discover this upload has been abandoned and reclaim the space that was allocated for the object. This trapped space will not be counted in your available space immediately, but it will after the cleanup is complete.  

See also: How do I fix trapped space?