what causes big clock drift?

A drift of 1-2 minutes is almost normal. 2 minutes out of 1440 is ~1%. Not really good, but not too bad (a normal PCs RTC is in the same range). I guess, for these, the basic PCB layout might be the reason. Parasitic capacitace will slow down the crystal. A common mistak is a GND layer below the crystal and specifically below the two wires that lead from the crystal to the MSP. YOu should not have a GND immediately below the HF signal wires, and any GND plane below the crystal should be separated from teh GND layer totally except for a small connection. The HF signals on the upper side (and current, that is) form low-impedance paths on the GND layer, channeling currents to flow below them with increased probability.

30 minutes or 3 hours is far off. Here the problem is more severe: crosstalk from a near signal line that is higher than on the 'working' samples (maybe even oscillation by an unlucky RCL or RC combination nearby, that coincidentally - by tolerance - forms a harmonic resonance with the crystal), defective PCB (micro-shortcuts that cause parasitic currents).

"HF follows strange paths" is a basic rule for HF engineers.

This could also be caused by a mismatch between the crystal used and the MSP430. What brand and model of crystal are you using?

Crystal = MIcroCrystal CM8V-T1A, ceramic package 32.768Khz supposed to be good for MSP430s.

Tolerance is +-100ppm, which if I'm doing the calculations right, should drift a max of 8 sec per day.

MIcroCrystal is very familiar with the needs of MSP430 and very responsive. I suggest that you contact them and tell them your problems.

Adam Hoover said:

Tolerance is +-100ppm

In my RTC applicaitons, I use a 5,6pF adjustable capacitor. It allows me to adjust the crystal frequency, so I get from several minutes to below 1 second drift per day. You see, a few pF make a big difference.

My next 30 units showed similar time drift problems.  Definitely a design flaw we need to fix, but I'm not sure whether to look at alternative crystals, use external caps (currently relying on the MSP430 internal caps only), or include a ground ring around the crystal (space is at a premium so we skipped the ring, but maybe that was a terrible idea?).

I've been trying to contact Micro Crystal but they are not responsive.  Anyone have a contact there?

Please disregard my last post.  I finally got to talk to someone at Micro Crystal and they were extremely nice and helpful.

Would you mind sharign the results with us (if any)?

Turned out we had 2 issues:  oscillator lines too close to LCD lines, and some oscillators being damaged during PCB placement.  The latter problem is being addressed in PCB manufacturing but the first problem was a design issue.  For those who see this problem, remember that LCD lines are square-wave alternating current which are very bad to have next to low-power oscillator lines.  For our next batch we redesigned the PCB to move them farther apart.  We also are trying to put a ground ring on an inner layer of our multi-layer PCB, to further isolate the oscillator.

I can't give enough thanks to the guys at Micro Crystal.  They were very helpful.  I shipped them a couple units and they tested the oscillators finding both our problems and making recommendations for future production.