Seemingly illogical pinout has been discussed before, but my question is different.
Have a look at the pinouts for the Si8540:
One would think that the pin order would be the same for both packages, maybe rotated, but that would be it. Instead the SOT-23 pinout is almost a mirror image of the SO-8's. Now SO-8 nor SOT-23 are flip-chip, so the die isn't mounted upside down on either, and I can't imagine that they would make a different die for the other package. I can only presume they crossed the bonding wires.
Looks like asking for trouble IMO, so why is this done? I can't think of another explanation.
Best Answer
Per Stevenh's recommendation, I am placing my comment as an answer and extending it.
Could be desired compatibility with a competing chip.. I used be in Semi and we have done this all the time. The idea was to offer a pin drop out and under cut them to get the biz. Wirebond is a sturdy piece of technology and last I checked people are moving to copper from gold. Most of the chips are tested at 50g drop tests, I don't remember us failing them at all.(BTW, these tests are never ever mentioned or released unless you are a serious customer due to hassle of answering questions)
In terms of signaling crossing wires is generally a no no, however don't assume there is a single pad for that pin. Depending on the marketing input, I remember we made several bond out options where we use different pads for the same functionality.