This website was created and deployed with terraform on AWS (CloudFront + S3).
A Lambda@Edge function gathers the request/response data, publishes them via the closest regional SNS. This should reduce the latency as much as possible. (Previous attempt was to talk to S3 directly, but that would hurt far away edge locations.)
A single region SQS is then subscribed to all those cross-region SNS topics, and triggers yet another ("regular") Lambda function to process the messages. This SNS->SQS bridge is very light-weight, not much to configure or to set up.
Finally the payload can then be stored in whatever data store we like (S3, RDS, ElastiCache, Keyspaces, DocumentDB, or even something external).
Even if message processing takes significant amount of time, this is now decoupled from the edge request-response life cycle.
Look forward to my TED talk and see me soon at an AMA of your favourite social media platform. ;-)