nx_arbiter.nx

nx_arbiter.nx -- ARBITRATION primitive for conflict-free PARALLEL workstreams (pillar 3 of the orchestration fix). Some sinks CANNOT be isolated -- they are inherently shared: the NAS push target (one wiki doc-root), _offc/ installs, the memory files (a sibling clobbered MEMORY.md + genesis_lineage.tsv mid-write today), the WMS queue, the git repo. For those, serialize concurrent writers with a KERNEL-ENFORCED advisory lock (flock) keyed by a named resource. A workstream takes the lock, does its critical write, releases; siblings contending on the SAME resource wait their turn. Uses sys_flock (nx_syscalls.nx, the RV64-32 -> x86_64-73 wrapper) -- NOT raw __syscall(73): the compiler translates rv64 numbers, so the x86_64 number passed raw lands on the WRONG syscall (proven: raw 73 -> EFAULT). Imports ONLY nx_syscalls.nx (NOT the legacy nx_flock.nx organ) to avoid the double-import rc=6 trap. fl_try(resource) -> fd>=0 if acquired now, -1 if busy/err (non-blocking). fl_acquire(resource, tries, ms) -> fd>=0 (bounded wait), -1 on timeout. fl_release(fd) -> unlock + close. Lock files: /home/.../.nishi/locks/<resource>.lock (shared BY DESIGN = the rendezvous point). license_tier: ORIGINAL