I don’t want to sound like a smart-ass (Or do I? 
) but you think wrong!
I found an excellent comment in a very, very fine written sketch about this and the use of setWaitForConversion(0); with 2 timers and 2 functions:
//============ Temperature sensor
/*
A note about about temperature readings from the DS18B20. The default resolution is 12 bits (0.0625 increments or 1/128 degrees C) and will take the sensor up to 750 ms to complete. This may seem like a problem, but it doesn't have to be (if I've done my homework correct)! startSensorConversation() tells the sensor(s) to start the temp conversation which would normally make the program halt for ~750 ms. But by setting setWaitForConversion(0) to false, the request just initiate the conversation and doesn't wait for it to finish. Total time for startSensorConversation() and getSensorData() functions is now a manageable 30-35 ms instead of 750-800! :)
The DS18B20 is rated for a minimum of 50.000 EEPROM Writes (but will probably do many many more). Even though it's a high number, getting a new reading once a minute equals to 1440 a day, and after just 35 days your over 50.000! So to be a bit conservative, I now only do it once every 10 minute. I'm still running getSensorData() once a minute so the data received is never older than 10 + 1 minute. That's OK for my needs.
*/
(Some spelling errors appears in the text; conversation should of course be conversion! )