True that many potential RCEs are found, but I think there are a few points to keep in mind.
RCE classification is often conservatively assumed when it is theoretically possible even if it is not been demonstrated. Android bulletins appear to assume any memory corruption could be an RCE.
Remote code is no longer sufficient for privileged control. Next, you have to use it to break out of a restrictive sandbox for whatever service or application you have compromised.
The baseband firmware is not so privileged anymore. Most new phones, like the Google Pixel 7, have IOMMU to force the baseband to communicate through a very restricted interface to the kernel. Certainly, you can interfere with texts and calls, but a baseband RCE doesn't yet compromise the data stored on the phone by itself--not to diminish the seriousness or to suggest that we shouldn't patch such an exploit immediately.
RCE, the "remote" aspect, in the operating system? So directly in the kernel and accessible remotely, such as through the networking code? I'm curious now. Most of the ones I've seen are in some other component that is sandboxed. True system-level privilege RCEs seem to be relatively rare. Usually, you get RCE, then you need privilege escalation to do something especially interesting.
Indeed; I'm sometimes able to leverage even a few bits of memory corruption into execution in many cases, though the hardened allocator in Android makes this a serious PITA to arrange to overwrite something useful.