BPF was our first alternative target we tried simply because a VM already existed. However, since then we made the switch to RISC-V. Our current opinion is that it is a much better target. Hence there is no need for BPF anymore. More information here: contracts: Support RISC-V bytecode · Issue #13640 · paritytech/substrate · GitHub
Implementation is already ongoing. It is still a long way to go, though.