I'm not sure this is the problem, but I've seen Audivolv crash on Macintosh and gNewSense-Linux because of slow graphics in version 0.1.6.
The graphics are only Java's standard drawLine function (with different colors and xor or not), but that is done many times per second. The target graphics level is is throttled depending on the predicted chance that the audio system had an error (like the sound skipping when the audio buffer gets empty unexpectedly fast), but on gNewSense-Linux (and maybe sometimes on Windows XP), it happens before audio starts playing. The graphics are drawn very slow, but this has also happened on Macintosh when the graphics were fast and sound was playing, and the only thing changed was going into some of the options parts of the window.
I'm almost sure many of Audivolv's event and thread problems are caused by thread errors, probably in the graphics or starting new threads, and possibly related to the DataLine objects as Audivolv chooses which sound-card(s) to use and display in the options list and/or just before the first sound is played or the first UI controls appear. There are thread errors, and they must be fixed in a provable way, not by "trial and error" programming.
Branched Audivolv 0.1.6 to version 0.1.7 and existing version 0.2.1 is not changed. Why? Because of these event problems. On gNewSense Linux, it the graphics were running very slow, causing the buttons to not appear and the sound to not start. Decreased minimum graphics level and let it auto adjust. The most important change was in how it selects sound card options. Java was saying it had lots of options it did not have, like 32 bit sound and 6 speakers. Audivolv was going through all of those first because Audivolv knows those are higher quality than the normal 2 speakers and 16 bit sound. Audivolv fails if its first choice for a sound option does not work, therefore in version 0.1.7 I told it to use exactly whats in the sound options window and not to let the operating system suggest possible options. It takes a few seconds to start the graphics and sound, but it works. This change should be copied into 0.2.2 when that version is created. There were a few variable names added that contain the word "force" in them near the audio choosing code. Also, in the Javassist part of the window, I added a code to only start playing sound if it is not already playing sound (when you click the button to use java code the user typed), which stopped it from going through the sound options again which was very slow or did not work at all. 0.1.7 works on gNewSense Linux but needs to be tested on other operating systems.