Since it seems the cache corrects itself, could Esri create a more public-facing error message so users don't think the surveys are broken? I know it is in the fine print but users only see the 404 error.
Also, it seems to be happening to quite a few people, is there a way to capture the errors and force cache updates automatically on the AWS Cloudfront servers to minimize the downtime? Short of that could a bulletin be put out to support so reports of these go directly to the S123 Webform/AGOL team to resolve it without having to go through all the troubleshooting questions prior to it being escalated. It would save a lot of time and effort.