The name is bacause I only ever considered making 20r,30r or 36r and I belive this will be beaten by an 20r solve. My 20r never materialized because anything I put on the board looked way too expensive.

Now let me explain the logic of the solve. You can skip it if alreadly apeared before.

If an input bit is fire, it will suppress some quicksilver and avoid certain projections.

In my head, I prefer to interpret it as salt adding some values to the output.

The encoding is the following:
ABCDE are the bits of the input XY are the seximal digits of the output
A will add 1 to X
B will add 1 to Y
C will add 2 to Y
D will add 2 to X
E will add 2 to X and Y unless C and D are salt, in which case it will add 4 to X

To implement this, we use arm 6 to check for (C or D) and adjust the placement of one of the fire atoms associated with E depending on the result.

The period is 36 because the quicksilver hexarm takes pseudoperiod 4 to unload the previous 6 atoms and pseudoperiod 3 for the rest of the projections. For an average of pseudoperiod 3.6

We can't 6P it to avoid the first unloading at pseudoperiod 4 because I paid just enough to get period 4 reloading.

Highlights (You may want to skip this and analize the machine, or look for a specific arm you want to analyze):
*Arms 5 and 6 reuse the logic debonder to deliver the conditional fire atom in a single swing and insert the non-conditional atom from E into the correct position.

*Arms 3 and 4 use the fire input as a conditional input.

*Arm 9 delivers the output as soon as possible while building the next string at the same time. This has a single restriction on which bits can affect which half of the string, but I decided to just always alternate.

*6P is used to allow arm 8 to deliver a quicksilver 1 cycle earlier without worrying about arm 10 grabbing one of its atoms.

*Arm 9 is in an spot where if needs to dodge some colisions from arm 8. This was an attempt to have arm 7 help arm 9 become a simple arm.