Some open questions:

• should we distinguish between production and dev modes?
• should we consider the network IDs?
• should the creation of the file path be dedicated to a function to deduplicate code a little bit?

The file path should definitely be differentiated by dev_mode and network ID. Especially in the case of local devnets being run.

In addition, snarkos clean should also remove these files.

Addressed the points in the previous comment.

This logic may need to be more comprehensive:

Given a scenario where the latest block is at round 50:

1. A node shuts down and stores its proposal at round 52 (before committing a new block).

2. On bootup the node will sync up with the ledger blocks and advance storage to round 50.

3. Then the node will try to load the proposal for round 52, but won't have the certificates in Storage to properly advance to round 52. Nor will it have the certificates used to create the proposal in round 52.

   • The node will need to request the certificates for round 51 from its peers and properly advance to round 52.

Rebased.

Closing in favor of https://github.com/AleoHQ/snarkOS/pull/3239