Although it is important to eliminate stress concentrations (nicks, burrs, tool marks, etc), doing so does not change a root cause of failure for aluminum parts exposed to cyclical stress.
Aluminum has no knee in its S-N curve. Given enough stress cycles aluminum parts eventually break, no matter how well prepared. This is not a problem if you can be sure the number of stress cycles in a lifetime will be low.
Here's the rub; anything attached to an engine can be exposed to a great number of cycles in a very few operating hours.
For example, assume a vibration at combustion event frequency. In this case (IO540) it would be a 3rd order vibration, or three cycles per crankshaft revolution. Assume 2400 RPM for cruise; 2400 x 3 x 60 minutes is 432,000 cycles in single flight hour.
The arbitrary number of cycles used to define the endurance limit of aluminum is usually 100 million. At 432,000 cycles per flight hour (in this grossly simplified explanation) the aluminum part is guaranteed to break in only 231 hours, even stressed at a very small percentage of its nominal strength. Higher stress, fully reversing stress, or stress concentration simply drives the number of hours to failure even lower.
So, please don't get the idea that "perfect" flaring and prep will somehow make aluminum lines immune to failure. It won't.
Here's a nice comparison S-N curve for steel vs aluminum:
http://www.ndt-ed.org/EducationResources/CommunityCollege/Materials/Mechanical/S-NFatigue.htm
Note how the curve for steel becomes a flat line by about 10^6 cycles. If stress amplitude is kept below the line the part will never fail due to fatigue. Now look at the aluminum curve and note it is still diving at 10^8 cycles....and it will continue that dive forever. This is why smart people use steel AN fittings on their engines.
My teacher called aluminum "frozen mush".