There is a perceptual problem with using linear fades when the end of the fade is reached. But there is no bug per se.

Here’s what’s going on:

On the meter side:
=============

Let’s assume we’ve got a 60 second wave file that is all fade-out (the fade-out starts right at the beginning).

Given a simple linear equation for the fade, the point where the fade reaches -60db is about 1/16th of a second before the end of the event. 1/16th of a second may be smaller than the meter update granularity on many machines.

On the perceptual side:
================

The dBs per second of attenuation increases exponentially near the end – and since loudness is perceived in dB space, the signal sounds like it is cutting out very rapidly at the end.

As a matter of personal preference, I much prefer Vegas’ smooth cubic fades (fourth from the top in the fade menu) or the log-curve (third from the top) when working with music

At the file level:
============

You can actually see that the fade is working correctly if you try to fade a DC wav over a sufficiently long duration – then examine the end of the wave file.