That’s what I was thinking, especially in the case of moving from 5 to 6. We don’t appear to hit similar issues on obs that often, otherwise?
Completely going by an hunch, and I don’t actually know what I’m talking about, but at least on aarch64 it’s (almost?) always __fxstatat64.
So my hunch/prediction is it has something to do with LFS (large file support), a.k.a -D_LARGEFILE64_SOURCE, -D_LARGEFILE_SOURCE, -D_FILE_OFFSET_BITS=64, a.k.a. -D_GNU_SOURCE, a.k.a #include <features.h>.
Now, akshually.
A typical FM is e.g. (WARNING) moc[5815/-143018176] Path not found for FD 3, for __fxstatat64()
The ‘Path not found’ comes from here, and we can see the third parameter is empty. Which means virtual_path must have been ‘\0’, and this jumps to the error handling. (Also the weird indentation hints at some merge error or so).
PR21 to scratchbox seems to deal with errors in this space.
Can someone test whether an sb2 binary with that applied fares better compiling Qt stuff?
I don’t think I can update the sb2 package from within an OBS project, right?
Unless we roll out our own OBS 
Hey! That was my idea!
And, in retrospective, maybe I should have tried to roll out some experimental OBS instead of rewriting tbuilder. It just seemed easier at that time… Anyone playing with private OBS already?
PS: sorry for off-topic. Maybe its better to discuss it under Qt6 for devs thread