* Run each offchain worker in a separate thread. * Start a thread only on supported targets. * Flip conditions. * Remove platform-specific code.