When starting with my 3i solves, I quickly through together a 240g/9630c solve, which I managed to get down to 210g/10038c within the first two hours. I reserved one spot on the wheel for each of the metals (not counting quicksilver). That solve looked likely to be my final solve, until I tried something new on the last day.

Here, I tried to only grab lead in 5 spots, meaning one spot needs to alternate producing two different metals. At first, this didn't work. But after shuffling around some of the outputs, it miraculously worked. (Thanks to the fast-forward function of FTSIGCTU, which I downloaded for that approach.) Thus, I had a 200g/16218c solve. Afterwards I shuffled the outputs around a bit more and settled with this 200g/9225c solve in the end. I might still have missed some shuffling around to save even more cycles, but at some point you have to stop.

As some others seem to be certain of having a winning solve and the 210g was easy to get, I'm looking forward to seeing how even cheaper solves might look like.

Analysis of the outputs: The top left spot outputs one Pb every loop. The right and left spot on the wheel get upgraded once every loop, but only get replaced by iron one round after being outputted. Sn (right spot) thus is outputted every 2 loops, Au (left spot) is outputted every 6 loops. The lower left spot thus is upgraded once every 6 loops and responsible for producing Ag. It thus needs 6*4 = 24 loops to produce each Ag.
The lower right spot is the most complicated. It is responsible to output Fe and Cu. It gets one Hg from the right spot every 2 loops and one Hg from the lower right spot every 24 loops. Usually it outputs Fe, but whenever these two cycles align AND there is currently Sn grabbed, then Fe gets skipped and instead Cu is produced and outputted. Every 24 loops there are 12 + 1 = 13 upgrades to distribute and luckily, the loops always align to produce Cu. So every 24 loops this spot produces 1 Cu and 5 Fe.