The full probe takes the console down
The minimal build boots healthy (previous entry). The full one does not, and this is the first time it has been through the install-and-launch path at all.
What was observed, and it is thin because the console stopped answering:
[Syscore App] createApp OBSC00003
[Syscore App] processSpawn(pid, 0xa018, ...) appParam: appType = 0, titleId = OBSC00003
[Syscene App] new pid=0x1dc attributeExe= 0
[Syscore App] Ready to exec
[Syscore App] wait for NOTE_EXIT|NOTE_EXEC
and then nothing. ps showed no eboot.bin; a
second launch attempt found shsrv gone, and
hw check reports every service down. The jailbreak needs
re-running before anything else can be measured.
No conclusion is available about why. The
log ends before EXEC and the crash-report block never
arrived, so it is not known whether the eboot loaded, whether it ran, or
whether it was the loading of 35,518 imports across 352 libraries that
did it. That is the whole reason pkg-min exists, and the
Makefile comment beside it says so: an eleven-megabyte filesystem and
four hundred imports has too many possible causes to reason about.
What is worth doing before the next attempt
The minimal build is the one that boots, and the distance between it and the full one is large and unbisected: four imports against 35,518, one library against 352, one section against thirty-four. The next step is not another full launch - it is something between the two.
EXCLUDE already exists for exactly this and names checks
or whole sections to leave unrun. A build with the census sections
excluded, or with BULK provably off (it is, on
HARDWARE=1, by D179's compile error) and only
base running, is the middle the bisect needs.
Also unresolved: /data is not writable from inside a
game sandbox, so the report sink the full probe uses has nowhere to
write on a console. The kernel log is the only channel proven to work,
and the minimal build reached it through
sceKernelDebugOutText.