There's quite a few reasons why this is never going to happen, in amongst x-validation. Your 'optimized' results wont validate against a results from any other random machine on the grid doing the wingman job.

There's the crux... this grid is contributed to by many large companies and colleges. It's very hard to convince these to participate here, much harder if WCG has no total control over the science applications. But anyway, you've gone on record... maybe the times they are a-changin.

Hello grid,

I bet all the mentioned companies and institutions use Linux, Apache or Samba to some extend, did the open approach of these make them any more insecure? Of course no college or other large entity would have to run "untrusted" compiled code from "someone on the internet". The optimized code could flow back to the code owner, must be reviewed and understood and if appropriate can be integrated to the master branch. If you don't like it, don't integrate it.
I wouldn't worry that much about the validation. Of course no one wants to waste compute time. Not the user of optimized apps, nor their wing men (which actually never loses, just waits longer for the resend to come back). Obviously those running apps through the app_info mechanism also want 100% revalidation successes as well. Of course there can be an initial time when results don't match yet, but with thorough testing this can be minimized. The numbers of invalid WUs and hence resends would be insignificant against what you need to resend every day due to failures from overclocking, disk crashes, lost devices, etc...

With so many active WCG projects, I'd understand that this or that one can't provide the source because they would have licensing issues or fear their code would be used by competing institutions or like that. But hey, is that really the case for all of them? After all quite a few BOINC projects are open these days, often run by universities as well.

Open code would give coders like N30dG the opportunity to provide applications to a widened user base. In my case I purchased 8 quad core ARM boards that I wanted to use for WCG. Back then I had no idea that the Android BOINC client is just a shadow of the real BOINC client, making a farm of Android devices (especially with Android > 5) hardly manageable remotely. So I'm using Linux on these, donating their time to Einstein@home and other projects that either provide ARM binaries themselves or release the code for others to create binaries. I'm convinced that Linux@ARM (not Android) is the next big thing. Scaling out with cheap ARM chips is perfect for todays big data, cloud and IOT business. Lets not ignore all that power (as is unfortunately done with GPU as well :-/ ). Lets be ready and have an app for these and other platforms, that have enough skilled users willing to port, optimize and contribute...

Yes, stumbled many times on these issues (not only excel, any spreadsheet programs), last week found HST1 was interpreted as a sheet cell (CEP2 is another), not a project moniker. IF you're aware it's not an issue and can code around, but...

if I was on his place, I'd contact the SRI (which is responsible for OET, OZ & FA@H, all ran on Androids) & offer their service to them!

but you can also go to SETi@home & improve their Android app!