We have experinced the same problem you describe but on 4 layer boards.  We used
permanganate desmear and direct metalization.  Although you say that the
interconnect failure looks like hole wall pull away, I think you will find that
the barrel is perfectly straight.  I'll also bet that if you measure the
resistance of the interconnect failure, you will probably find something in the
20-2,000 ohm range.  That is how we positively located the interconnect failure
sites.  In some instances we found that the "gap" between the innerlayer copper
and the hole wall copper to be almost 0.25 mils.  Even with gaps as large as
this we have measured the resistance as low as 20 ohms.  Our continuity
threshold is 10 ohms for board testing.

We feel very confident that the "gap" is really the direct metalization
material.  You have probably noticed that your cross-sectioning polishing cloth
is becoming very dark.   It is there because there is a trace amount of resin
smear  is on the innerlayer copper.  Many of the direct metalization materials
that I am familiar with have a tremendous affinity for resin.  If there is any
resin smear on the copper, the direction metalization material will clump onto
it.  Since the microetch cannot penetrate the smear, it stays there.  The
junction between the inner-layer copper and the hole wall copper is the
metalization material.  Although they are good conductors, they do not hold up
well under HASL and wave soldering temperatures.  Wave soldering is worse
because of the amount of thermal mass in the hole.  We think this is why the
innerconnect failure will survive HASL and not wave soldering.

In fairness to the direct metalization manufacturers, I don't think this is flaw
in the direct metalization process itself.  I know of one board shop that
produces well over 750 multlayer panels per day using this process and has not
had a single inner-layer failure as described above.  


***************************************************************************
* TechNet mail list is provided as a service by IPC using SmartList v3.05 *
***************************************************************************
* To unsubscribe from this list at any time, send a message to:           *
* [log in to unmask] with <subject: unsubscribe> and no text.        *
***************************************************************************