Sorry about that! I can see from the logs on my side that between 1:44pm and 3pm the cloud environment in the US where I host ML for Kids was very unstable, and had eight outages (Fortunately each one was reasonably brief)
For the Machine Learning for Kids website and training tool, I have DNS failover set up which means when the cloud environment I use in the US experiences problems, the site automatically switches over to an alternate cloud environment in Germany - so avoids these sort of outages. [1]
However I don't have the same multi-region setup for Scratch - so when the US cloud region has problems, it results in the error that you describe above.
I'm very sorry for any inconvenience, and I'll look into extending the multi-cloud setup that I have to include Scratch to avoid this sort of incident in future.
Kind regards
D