I have seen that claim too, but I am pretty sure it is just an speculative urban myth. As explained, every 14 mm DT Swiss has an ideal spoke termination length 1 mm below the top of the nipple (just like an 12 mm nipple with 1 mm too short spokes), I have never heard of anything like stress cracks in 14 mm nipples. Considering that Mavic recommend 14 mm nipples for their popular and widely used CXP 33 rim, there should be loads of problems with that rim, but I have never seen anything. Been using that rim for almost a decade on several wheelset, and never had a problem, even when using +140kgf spoke tension.
I also can't visualise what forces that could wiggle a spoke inside a nipple on a tensioned wheel; either just the spoke would flex from lateral stress, or the nipple and spoke would move together. So I don't believe in neither stress points at the spoke end, nor in nipples popping off because they only had 7 mm instead of 9 mm thread engagement.
Anyway, 16 mm would give full 9 mm thread engagement for spokes 2 mm too short, and saves a relacing and new spokes.
Roger Musson, author of the instructive "The Professional Guide to Wheel Building" has instructions on how to measure ERD and build the tool to do it:
Check this pdf from chapter eight from his book:
http://www.wheelpro.co.uk/spokecalc/calc-measure.pdf
(IMHO his book is worth buying too)
Basically you take two spokes, cut them so they are 200 mm long each. Glue on a nipple on each so the spoke end are flush with the top of the nipple.
You put each "measuring spoke" in opposite spoke holes in the rim, and use a ruler to measure the the distance between the two spokes.
The entire distance between the two spoke ends are the ERD (2x200 mm+ruler distance=ERD)
Put that measured ERD number into Jobst Brandt equation, or a program that uses it like "spocalc", and you will get a correct "raw" spoke length. Round it down 0.3-0.5mm to compensate for spoke elongation and rim compression.