It's not necessarily you doing something wrong, this might be up to Google or GRC Pool as well.
I didn't get this message ever before, but I just got it now upon checking. I've done a little bit of testing and didn't manage to figure out easy solution, however I noticed that my VM was still able to comunicate with GRCPool with use of boinctui (it's later on in the guide), send and receive tasks and synchronize with saved manager. While if I tried again with boinccmd -join_acct_mgr command it was either unable to find localhost and on some settings unable to Authorize like in your case.

You're luckily not loosing much because there's no risk of loss of MAG if you didn't have any in the first place.
All my other VMs that have been running for around a week now are still able to communicate with the pool.
I didn't check if boinccmd fails to join the manager.

I deleted my instance
Google Cloud Dashboard > "Go to the Computer Engine dashboard" on the bottom of Computer Engine box and chart >Click on 3 dots in last column of your instance row and choose "Delete".

In the meantime go back to GRCPool, go to your dashboard, click on hosts box. If you don't have any hosts here that's fine you can close the window for now. If you see your old VM-instance host on the list, click red button below that enables host deletion, then click the arrow by the end of the row with the host name and click >Delete Host and confirm with "Ok"

Go back to your VM instances list, and click that small blue icon with + inside to create a new instance.
From this point you can return to the guide and continue from the instance setup.