If this posting generates any interest in this subject matter, I will be happy to go into greater detail, but for now I am going to take the 30K' view...
Beginning about 20-years ago, we studied all returned leaky connections for a period of five years...our connectors and those made by others. (We supplied 75% of the world at the time, so most assumed any connector was ours.)
Typical quartz quick connects (aka press fit unions and splitters) only seal on about 0.1mm of the column end or so. Three problems cause 90% of all leaks. If there is a chip in the column end larger than about 0.1mm, it'll leak. If the polymer coating (polyimide) is torn more than about 0.1mm in scribing the column to make the snap cut, the column will probably seal until its thermally cycled. If the column is cut at an angle where the face encompasses more than 0.1mm of column length, it can't seal.
Two related problems with the capillary used to make columns cause about 2/3rds of the remaining leaks: oval capillary and a lack of concentricity of the coating. Where these two problems overlap, leaks are virtually assured (thin polyimide on the short diameter of the oval capillary). All capillary is oval to some extent and the polyimide coating thickness can vary by as much as 50% about the diameter of the silica tube. Where ovality or concentricity is high, either one, alone, can cause leaks, but there is usually a contribution by both problems. In fact, both these problems exacerbate the first three listed as well.
The final problem that causes leaks in cracked capillary. This problem is rare, but far more common that one might expect. We're not talking about circumferential cracking -- that is a problem I lump in with large chips -- but axial cracking. Some capillary is drawn too cold and high levels of axial stress remain in the glass as a result. The zone of plasticity where capillary can be drawn is very narrow: too hot and it pours out of the furnace and too cold and it won't draw down at all. For this reason, and for the complementary fact that time is money and faster is cheaper when drawing glass on a $1M machine, manufacturers err to the cold and fast end of the production window.
It is likely that the reader has noticed that no problems with the connector have been identified in this discussion. Suspicious? Not really. All except the last reason for leaks can be accommodated by good connector design, so in fact, 99% of leaks are the fault of the connector because the realities of the sealing condition were not fully addressed. At my company, we have now changed that.