I think they'll make incremental progress at accepting more programs, but that BPF-C programs aren't going to look like normal C programs any time in the near future.
There's a bunch of data structures available to BPF programs, because they can call "helpers", which are just ordinary C functions. There are concurrent semi-durable key-value stores of varying types (eBPF "maps"). These aren't written in BPF-C; they're just kernel code that BPF gets to call.
Adding new helpers is not nontrivial; you can't just load them into your kernel and call them; they're part of the kernel API.
There's a bunch of data structures available to BPF programs, because they can call "helpers", which are just ordinary C functions. There are concurrent semi-durable key-value stores of varying types (eBPF "maps"). These aren't written in BPF-C; they're just kernel code that BPF gets to call.
Adding new helpers is not nontrivial; you can't just load them into your kernel and call them; they're part of the kernel API.