Did you try to dump the state of all the threads? We’ve seen issues in the past where the IO dispatcher was starved on decompression tasks because all threads were busy with PSO decompression at startup, and the loader was stuck on waiting for IO because nothing was making progress.