The underlying issue was resolved by CloudFlare. We believe it was related to this incident:

https://www.cloudflarestatus.com/incidents/npdcw7836xjq

We've removed our workaround to ensure that images are back to loading as quickly as possible.

We've added a workaround to the CloudFlare fault which has largely resolved the broken images but has reduced the overall performance of image rendering. We'll wait for CF to release a fix and then revert back to our old mechanism asap.

We've narrowed this down to a problem with CloudFlare which is failing to proxy some requests back to the asset storage platform. We're currently trying to escalate this issue with them as we've seen that it is impacting multiple CloudFlare customers.

Image download speeds are largely back to normal. We'll continue to monitor

We're seeing slow image downloads for some files which are not cached within the CDN. We've reached out to our storage provider to investigate further.