What happens is when Express Talk puts a call on hold, it signals to the VoIP provider that the call is on hold (rather than just muting the call and playing the 'music').
Most VoIP providers will ignore this on-hold command and then Express talk can continue to pass whatever audio it wants (this is what happens in most cases, and thus callers hear the Express Talk 'music')
In other cases, the VoIP provider will receive the on-hold command and decide to put the call into its own holding queue (what appears to be the case here). In some cases this just leaves the caller with silence, and in other cases the VoIP provider plays its own music.
This issue has come up several times, and I think it's time NCH add a "disable on-hold signalling" option. I think the reason ET does signal is due to the fact that it was built to connect with Axon which uses the signalling to play the music/audio stream set-up on the PBX side of things, independent of the local settings.