This was exposed by sqlite3 updates to 3.50.x. A way to reproduce and observe the issue is described in this message: https://lists.openembedded.org/g/openembedded-core/message/222660 (see also the thread for context)
Ross thought this was a high.
Discussed on the bug triage call 2025-10-09. https://git.yoctoproject.org/poky/commit/?id=84bd11c62e860031c1615ada76d7140b4717276a is probably relevant to this. Need to look at if this is needed again or if we need to handle pseudo-native:do_populate_sysroot_setscene specially. Potentially copy sqlite shared library into pseudo-native and add RPATH so it is always found? This is horrible though.
Patch sent to the list: https://lore.kernel.org/openembedded-core/20251016-fix-pseudo-native-v1-1-7c42af094122@pbarker.dev/ This looks to do the right thing in local testing :)
I can confirm that the setscene sequence becomes correct with the patch.
The issue tracked by this bug is now resolved. Further improvement of hard setscene task dependencies will be tracked in bug #16097.