It looked like no goroutines were getting scheduled. Even our internal debug http server was not responding, which I've never seen before.
I think I have been able to reproduce it locally on Mac and linux (docker). On Mac it seems to use 50% CPU instead of 100%. Unfortunately the repro involves running queries against the cassandra database using the gocql driver.
Get cassandra running docker run -p 9042:9042/tcp --name cassandra-repro -d cassandra:latest
Install gocql go get 'email@example.com' (didn't try other versions of gocql yet).