This question has been asked before, but none of the answers really seemed conclusive so I thought that I would bring the subject up again.
Here is the scenario:
I am a reseller hosting with one company.
I have ns.mynsdomain.com and ns2.mynsdomain.com.
Each of my client's domain records lists my nameservers (as opposed to the actual host's dns)
I have one client in particular that is more or less mission critical. So, I want to open a second account with a sererate host with a different NOC and use ns3mynsdomin.com and ns4.mynsdomain.com. The e-mail issue I would assume would be minor since this clients e-mail is mx'd over to a third party mail server.
By doing this can i increase uptime and insure against failure? I am assuming that if ns1 and ns2 crash and burn that ns3 and ns4 can pick up the slack.
Any comments, references or suggestions would be helpful.

If the services are mission critical they should deploy load balancing, global server load balancing, or some other technology that is designed to compliment mission critical hosting. MacGyvered DNS is not the answer.

To answer your question: No, doing this will not increase your uptime. Increasing the number of DNS servers does not have any correlation to uptime, what it MAY increase is the availability of your website. There are some caveats that need to be considered though, the first of which, is that you can't control the TTL other ISPs set their caching name servers to. In order for this DNS idea to work, you have to have low TTLs. If an ISP does not honor your low TTLs, then it will not work as expected.

Also, you will not be maintaining any sort of session information between the servers. If a server does die, any information on that server (shopping cart, etc) will be lost.

Finally, even if you set your TTTL to 1 minute, and you have a customer who is on a server that crashes, they will have to close their browser and reopen it before they can get back to your site, because Microsoft has incorporated its own mini-DNS server into IE.