I note that VMTest of the affected boards doesn't fail. So will a vm-test pre-cq be able to catch such bugs?

AFAIK we are not planning to run ARM VMs for performance reasons. This leaves the proposal at adding a 32 bit Intel userspace, which is not used in any product. The intent of the VM is to get some cheap/scalable but somewhat dirty coverage. I am worried that supporting a non-product makes this an expensive proposal. I would accept this failure as a limitation (dirty) part of the VM testing approach.

That's fair. I'm just usually worried about things that don't fail in Pre-CQ but fail in the CQ.

To Xixuan's point, this was a system call that was succeeding in 64-bit userspaces and failed in 32-bit userspaces -- we would need to run 32-bit VMTests for things to fail.

Re#3: I agree with ihf@ here 
- arm VMs are too slow for anything usable.
- We should probably re-enable a 32bit VMTest board both in pre-cq and cq (we should avoid pre-cq targets that don't have a cq equivalent, because they're more likely to break)

We don't have working support for vmtest on arm platforms. This is probably a good idea, but we need product support for this.

This bug is very old, is Untriaged, and has no owner.  If it is still relevant, reopen as Untriaged or open a new bug