"toll quality" basically means that whatever path your voice takes thru operator's network, QoS, CoS or whatever the service management method applied to your frames, package, signal etc. the end result would be identical to the performance of a legacy twisted pair cable that phone systems used to be built on. A toll quality connection as a general rule:

1) Would not require any specific equipment other than a native phone device (assuming that PRI ports are native to the phone system, which is a subject of never ending discussions amongst some old farts like me around here...)

2) Would not cause any digital disturbance to voice quality like packet loss, jitter etc.

This reminded me a story from ten years ago:

A world famous telecommunications equipment provider (a company from North America with a rhyming name :) ) claimed to our management suits that their products are five nines, including maintenance window (!). Naturally after a short while in our (non North American) local conditions they started to pop like corn in a hot oven. Their explanation was (and this was somehow implied in colored pieces of papers they provided to upper echelons of our management) that their power bus (provided that supplies connected to different fuses, preferably of different phases of line) is five nines, thus the box is always online, regardless of the status of CPU :) We started to change CPUs, then supplies of boxes started to pop as well. The new explanation was although it was the recommended practice to connect PSs to different phases, the actual frequency shift in our local conditions (50 Hz 220 V) was far above the conditions in their R&D environment (60 Hz 110 V), thus our local level of variations has a greater absolute value than expected. Shocking, mentally and physically as 90 V respective voltage to the real ground manifested on chassis :)

What I am trying to tell, as an old fart, is that redundancy starts at the bare metal level. Any design on an upper level must be supported by the hardware. Also it is important to know what is being made redundant. You can have multiple Internet connections but if your UPS is not redundant and is located near brown water facilities, you can find yourself in deep shit...

Even our crappy (pseudo-)government run telco offers 1-N real-time backup for E1 trunks, they even provide geographical redundancy (with no surcharge), if you are willing to distribute your ingress... All CLECs on the other hand are offering SIP trunks, with all the bells and balls attached... Albeit it is a crappy service in general, we prefer incumbent operator's E1 based service, because they know what "toll quality" really means... Moral of the story, if a telco cannot offer redundant E1 or SIP trunks you need to consider changing operator, before starting to think about servers...

Then neither you are not working with web developers who change their mind daily about basic system design nor you are working in an environment where uptime is not a concern nor you have very expensive load balance boxen that protects you from mundane facts of life. Either way, you are a very lucky person... OTOH we, of lesser levels of humanity, are grateful for one more tool that would make our lives easier.