Postpone implementation of race conditions safeguards for existence checks

Related to #300 (closed).

This MR revisits the conversation from !373 (comment 437658636) (more precisely do we really need to fix this?) and proposes that we do not implement race condition safeguards around blob existence checks, at least for now. Although the identified solution is technically feasible, the cons clearly seem to outweigh the pros.

Please see the diff for the rationale. In short, given that these don't represent an integrity risk and that, likely, they would only occur rarely, we should approach this iteratively and avoid increasing the risk around DB locks until we're certain this is necessary.