This destroys the atomicity of pwrite() and thus suggests that we should do something else to ensure
that multithread access to peakfiles (if it does in fact exist) is safe. OTOH, there may be no
such thing, since they are designed for the GUI, which is single threaded (for now)