I still can't believe  it. I am pretty used to optimising for rate, but actually caring about latency makes everything SO MUCH more difficult.

At first I made a layout and saw that it was close to a 25, but with no chance of a 24. So I moved on and remade it into one which had a possibility of getting 24, and oh my god. I spent hours and hours restarting from that layout. I already gave up on 24, so I was willing to lose a tick, but it always just barely didn't work. Either a very close atom collision stopped me, or the impossibility of getting the last quicksilver to the bonder in one tick. Finally I came back to my very first layout and saw that it was not very far from actually working. A few hours later and I had the 25. Some cost optimisations resulted in the solution you see here.

If you stay at max rate, this solve is min+2 which is really good I think. One tick is wasted on regrabbing the metal and in parallel sliding the quintessence onto the bonder, and another is wasted on sliding the output at the end. I am sure avoiding the final slide is possible and we will see a 24.

23 at max rate would require no metal regrab, which I think is questionable, but this community never fails to surprise me.

Of course the reason that I keep saying "at max rate" is that you don't need to make more than 6 outputs. That means that the 7th input is free for the taking: you can use parts of it (an element or a quicksilver) to speed up the making of the 6th output. Perhaps you could gain one or two ticks with that method, however it is beyond my ability.

Sorry for the rambly notes. Now let's see the top solves! So exciting