@funyjoker, while non-homogenous node configurations are supported, they are not advisable and you need to fully understand the implications of configuring you cluster in a manner that is not recommended.
The memory part is fairly trivial - it just means you won't be able to run as many VMs or less VMs with higher memory requirements on those nodes. That may also have implications for HA and you might not be able to restart all VMs in some situations if the remaining nodes don't have sufficient memory.
The storage aspect is IMO worse - those new nodes will be pushing 2x the throughout compared to original nodes and that could have implications for other aspects such as NIC(s)/switchports becoming a bottleneck on those nodes. It would also have impacts for repairs from failures and data evacuation - you would be far better off just adding 1x 4TB disk to all nodes so they all have 3x.