Is the "best" solution here focused on zero downtime or just minimizing impact? If strict zero downtime is required, would option D (high availability) actually be safer? Docs and exam dumps seem to differ here.
Yeah, D makes sense since using an external replica lets you sync changes and do a quick cutover. Full dump/import like in B would mean much more downtime. Pretty sure this is the right Google-recommended approach unless there's a version mismatch.
For this type of scenario, you'd usually go with A and E. If the corruption is big, the backup/restore process is necessary, but for minor or isolated issues, you can do a stale read at a safe timestamp and just overwrite those rows. If it wasn't asking about extent and wanted full restore every time, then D would be worth considering, but that's not what it says here. Pretty sure about this split-any objections?