It's the hold music. VOIP bandwidth is dynamic. When there is no voice (or sound) passing over the "connection" then the bandwidth that had been carrying the voice (Muzak) information is available to carry other information (i.e., data). In systems designed to give voice traffic priority (aka QOS meaning Quality Of Service such as 802.1p/q) enough VOIP traffic can bring "surfing" literally to a standstill. However, I know of no such VOIP system that supports QOS across the public Internet, so worst case the VOIP traffic will simply be competing for the same pool of bandwidth that your surfing is competing for. That of course means than not only can "Muzak" (or an actual conversation) slow down surfing, but surfing can also destroy the quality of the call, making it unintelligible gibberish.