Make error handling more consistent

The way that errors are handle across the platform is inconsistent, e.g some commands are multiple ops where some can fail but the batch completes, while others will completely fail if one part of the operation does.

No files, tests, or specific command entry points are named. Start by mapping how commands handle multiple operations and failures, comparing the partial-failure and complete-failure behaviors described in the issue. Done means the project has an agreed, consistently applied error-handling model across commands.

Written by the indexing model from the issue text.