I used a binary-ish encoding of the input stick and encode to a big-endian base-6. It's binary-ish because the largest digit isn't the usual 16 but 18 instead. This takes advantage of the fact that input doesn't map 'onto' the output space, there's some wiggle room. Not much, but it's there. Why do this? It simplifies the carry in the upcoming adder.

I use triplex bonds to accumulate sticks of quicksilver to send to the adder. But first another trick: With this encoding, neither the smallest nor the largest digit affect the need to carry. So they can skip the adder altogether and just build the base for the quicksilvers to project. This leads to a somewhat awkward alternating smallest/largest digit placement in the input, it's neither big-endian or small-endian, but mid-endian with the middle-most digit at the end.

The adder builds up the corresponding quicksilver sticks then does a quick test for length 6+. If it's 6 or longer we need to carry a quicksilver over to the larger digit. The binary-ish encoding means we never need to worry about carrying more than a single quicksilver, there's not enough '1s' to overflow to 12+.

The rest is really just trash management. The disposal glyph gets a workout here, the timings worked out pretty nicely in the end. Somehow they all just miss each other with only an occasional stutter needed. 

Great tournament this year, lots of fun puzzles! Congrats to all those who won rounds, the top solves are always far above and beyond what I can imagine! This has been a Symposium to remember, enjoy the banquet and drinks!