There are two problems with your statement. First, it's unclear whether what you save pays for what you gain. Second, the participation of memory unsafety in vulnerabilities is not binary. If you can cheaply reduce memory unsafety by 95%, that doesn't mean that paying a high price for the remaining 5% is worth it. Yes, memory safety is the single greatest source of security vulnerabilities in some languages (it is not in general, but then again most software is already written in memory-safe languages), but once you've already reduced it by 95% it no longer is. (BTW, I'm not even sure that soundly eliminating the 95% is clearly worth it compared to alternatives, but that's a separate discussion).
There's this obsession with soundness that is undoubtedly interesting to researchers of sound methods, but isn't actually justified by the real-world needs of software (sometimes it is and sometimes it isn't, but when and by how much is something that can only be answered empirically; it isn't something that can be simply asserted). Software correctness is at least as much art as it is a science, and ignoring the empirical aspects that make it into an art and focusing only on the aspects that are science because that's where things are clear-cut doesn't make those aspects disappear. At the end of the day someone has a budget and they need to spend it in the way that yields to most correct software; how that is best done is an empirical question. After all, you started with an empirical finding about memory safety being a leading source of some bugs, but it simply doesn't follow that total memory safety soundness is the answer to the empirical question that we need answered. Soundness is clear-cut and mathematical (and that's also why it's attractive [1]); sadly, software correctness is the very opposite.
We must continue researching sound methods, but if anyone thinks -- knowing what we know now -- that this should be the main focus and the best answer to software correctness then I think they are misjudging the nature of the problem.
[1]: Ironically, people who focus on soundness love proofs, yet they try to skirt around the proofs of the futility of soundness by resorting to empirical -- though untested -- claims.