What it really boiled down to is that some routing logic was needed and doing it at the edge seemed like a simpler solution.
In one case a site was moved from a sub-domain to a sub-directory of a domain using a reverse proxy. What was needed was a way to determine where a request was coming from to route the requests appropriately. Using edge compute, we could have looked for a header from the proxy, but on the server, the header was not reliably there in the case that a request had been already been served from the cache.
The other use case was to do some routing.
In both cases, a decent solution was found, but having something more integrated into the hosting platform would have been a little nicer.
Hopefully, that helps, but let me know if I can clarify anything.