Hi… If this is the legacy Fly Postgres (as opposed to the newer managed postgres) then the first step is to simply check its logs with fly logs -a db-app-name to make sure it isn’t OOMing, running out of disk space, etc.
If not, then it may just be a network limitation. A fellow user was reporting problems with a similarly sized transfer recently. Unfortunately, long-running TCP sessions are prone to getting cut pretty much anywhere along the path through the Internet, 
.
I think the Postgres wire protocol was designed for LANs, originally, but I could be wrong there.
A more reliable approach is to use an Internet-oriented transfer mechanism with compression, like gzip + SFTP or rsync, and then do the pg_restore within an SSH session on the database Machine itself.
Hope this helps a little!