Replacing broken hardware for free is a generous offer. However, as you said in Zencuke’s omniwheel-robot topic, it is probably best to get more information about the problem first. At least 3 people have similar problems (probably 4, because I think Ivar’s problem of a rebooting RPi when running the motor example is related). It might therefore not just be a case of bad SN754410 driver ICs. The topic about the ultrasonic sensor returning wrong distance values showed that there probably are different versions of that sensor around. This made me wonder if that could also be the case for the NXT motors.
Searching for information about NXT motors I found this very nice webpage:
It provides lots of information about the performance of the motor, but also about its internals.
A nice schematic can be found at:
Note that these two sources disagree on the type of PTC resettable fuse that is used in the NXT motor.
If the PTC is indeed the RXE065 or MF-R065, as the first link suggests, then at room temperature the fuse would trip at a sustained current of 1.3A (see spec.sheet at: http://www.philohome.com/motors/RXE065.pdf). This is well above the max. allowable sustained current of 1.1A of the SN754410NE (spec.sheet at: http://www.ti.com/lit/ds/symlink/sn754410.pdf). Moreover, at room temperature it takes this PTC 5.3seconds of 3.25A surge current to trip. Again this is well above the allowable 5milli-second 2A surge current of the SN754410.
If the PTC is the MF-R030, as the schematic of the second link suggests, things look a bit better. The maximum sustained current at room temperature then is 0.3A and it trips at 0.6A, all within SN754410 specs. However, also this PTC needs 3 seconds to trip when handling a surge of 1.5A at room temperature.
The question now is if such large current surges can happen. I think they can and that the condition of rapidly changing rotation direction at full power is one of them.
If the above supposition is true, then for the case the NXT motors have the MF-R030 PTC, the solution could be to stack two SN754410 ICs, like shown on this webpage: http://letsmakerobots.com/node/10509
If the NXT motor has a MF-R065, well then I guess another way to limit the surge current needs to be considered.
In the mean time it may be wise to limit the number of AA batteries in the pack to 6. I noticed that a fully charged 8 pack has a static voltage of 11.5V. This will drop when loaded of course, but with 10V it is still above NXT specs.