To raise an analogous problem: a DOS system starts by loading device drivers in config.sys and then executing commands from the autoexec.bat startup file. It doesn't keep a record of error messages that occur during the loading of device drivers, and often they do not remain on the screen for long enough to be read when, in autoexec.bat, a screen reader is loaded. By contrast, both Linux and Emacspeak maintain records of their error messages: Linux in a log file and Emacspeak in the *messages* buffer. Thus, once your auditory interface starts, you can read these records -- a situation which is significantly better than the DOS example outlined above.