Based on similar issues filed and already closed without resolution, I found that the function _lock in bin/internal/shared.sh presumes well behaved flock functions in Linux. However, at least Centos with NFS (my setup), the presumption is false, and the function fails. It works if patched like this:

function _lock () {
  # if hash flock 2>/dev/null; then
  #   flock --nonblock --exclusive 7 2>/dev/null
  # elif hash shlock 2>/dev/null; then
  #   shlock -f "$1" -p $$
  # else
  #   mkdir "$1" 2>/dev/null
  # fi
  mkdir "$1" 2>/dev/null
}

But that clearly is not a universal patch.