For a cargo ship, does anyone have to be on board at all? Why not control it remotely from a smaller escort ship that's proven safe and can easily and quickly move away. The cargo ship's captain could sit in a simulated cockpit on the escort ship (the escort ship having its own captain). Just have the crew in a separate, safe, more maneuverable ship.
This is the $10 billion question running in maritime circles right now. There's actually no real need or benefit for the watchstanders to be on a nearby boat--that probably doesn't save time or money. For most of the trip across the ocean, control by satellite would be fine as long as you could get bandwidth to transmit radar data up to a land-based control station. With constellations like Starlink, that's much more possible than in years past. In most cases, the larger ship is safer, even if it is less maneuverable.
For 99.99%+ of transits, everything would be fine. It's handling that other fraction when things go wrong that is the issue. Think of how bad it got when the very small error rate hit the Suez Canal and the Ever Given got stuck. Some other issues:
* Every so often, things go badly wrong. A ship catches fire, steering goes out, engine breaks down, etc. You are far, far more likely to save the ship and cargo if you have a crew on board.
* The crew does a significant amount of maintenance when they're on board. For example, virtually all commercial boats have enough redundancy in their generators to have one down for maintenance. The crew can do nearly any maintenance needed by taking that genset down and just using the others. Multiply that by winches, deck equipment, painting, etc. and there's a lot that happens at sea. It would be hard to shoehorn that into maintenance during a port visit.
* You will need a crew on board when you dock to handle mooring lines and manage machinery, not to mention having a harbor pilot on board. Getting them on and off is difficult and dangerous, particularly if you have to do that offshore.
* There are some things that people do better that machines. There's pattern recognition, and sometimes a gut feeling that something's not quite right. That can be as small as an increase in the roll period indicating flooding in a compartment. While that data can be captured in an automation regime, it's hard to replace the feeling you get for a boat and how it normally behaves.
As you probably saw in the self-driving cars thread, I'm somewhere between a skeptic and a pessimist on the promise of automation to replace people, at least with current technology. There are definitely answers to all of these concerns, but self-driving ships are still in the very small boat demonstrations phase.