Please correct me if I'm wrong, but I think it's safer to use `EXTRADLLFLAGS` for specific modules instead.
I don't know about safer. If anything I'd argue this is the "safer" solution in a sense, since it turns out that more than just d3d tests are running out of address space, and I don't want to spend ages playing whackamole. By contrast there's only a couple of tests in ntdll / kernel32 that make assumptions about the address space limit.
It turns out that this also addresses bug 55725 (amstream:amstream), and I wouldn't be surprised if also addressed bug 55605 (winmm:mci).
Otherwise, could you take a look at the following failures?
Addressed, plus another test failure in kernel32:heap.