Hi @Baudota, welcome to the community and please accept my sincere apologies for the trouble this causes.
A 503 error typically suggests that n8n isn’t running (which in most scenarios means it’s currently restarting).
From looking at your question it seems you might be using n8n cloud which unfortunately only has a very limited amount of memory available in the smallest plan. So, it seems likely that your workflow execution consumes more memory than available to your n8n instance.
Perhaps you might want to consider moving the logic your loop (fetching HTTP data, processing it and adding it to a Google Sheet) into a sub-workflow? The sub-workflow could then return only a very small item back to the parent, and thus freeing the memory required for each batch. A recent example for such an approach would be over here.