Eric,
Susan Mansilla had some good comments regarding process qualification:

a.  What does your customer need to see?
b.  What testing did you do to ensure that the hardware you build now is
good, then repeat that testing using the new process.

You indicated that you use MIL-STD-2000 and J-STD-001.  Both have a process
qualification protocol.  MIL-STD-2000A has Appendices A and C.  J-STD-001B
has Appendix D.

In my experience, many customers don't know what to specify in terms of
qualification data, and it often turns into the Blind leading the Blind.
Second, when I ask clients what testing they have done to ensure that the
current hardware is good, they often don't know what testing was done, or no
testing was done, or the test data has been lost in antiquity (e.g.
retirement of the local guru).

In most cases, the client is just looking for some assurance that you have
done your homework and that their hardware will work as functioned for as
long as desired.  In my opinion, neither of the above protocols will do this.
 Both rely on SIR testing and CAN catch tendencies for electrochemical
failures, but there is often a very big leap from the IPC-B-36 test boards
used for the qualification testing, and actual hardware.  In my opinion,
burn-in or life testing of actual hardware in hot and humid conditions is of
more use than SIR data (heresy, I know), since the link to the hardware is
much more immediate.

Many of my defense contractor clients have gone the route of MIL-STD-2000A
qualification as the path of least resistance.  Do the test, pass the
minimums and you can use the process without buy-in from the customer.  They
then use the grandfather clause found in paragraph 4.2 in J-STD-001B to say
it qualifies to that specification.  I have had only a few clients brave
enough to attempt the J-STD-001B, Appendix D protocols, since there are so
many unanswered questions in that protocol.  What some defense clients have
done is to qualify the process to the A revision of J-STD-001, which is much
easier and lower in sample size (3 vs. 10), and then grandfather in.

I have one current client who manufactures Class 1 material worldwide.  They
were initially going to use J-STD-001, but it was a political decision, not a
technological one.  Rather than use an SIR test which had a dubious link to
thier processes, they based a protocol on ion chromatography, modified-ROSE,
ROSE, and their own internal life tests using temperature and humidity
cycling.  This is what they currently use for separating good from bad
product.  The ion chromatography data is being used to show what residue
levels correlate to good and bad performance.  The ROSE and mod-ROSE tests
are being done to develop adequate process control guidelines.  Their
life-tests are being used to determine good and bad.

Hope this helped answer your questions.

Doug Pauls
Technical Director
CSL