I think I did well on this puzzle so I've written a lot of notes.  If I place at like median or something you can skip a bunch of these.  Skip whichever ones you've talked about on previous solutions as well.

The main constraint for a 19A solution is that you'll end up with 12 waste on the board for the final steps.  I didn’t know if the best solution would have just the red reagent or both until I made the final layout decision.  The waste is 12 either way, so the limiting factor was how I make the output and give myself space to work.  I found that it was helpful to think about the empty spaces rather than the spaces with waste.  There are 2 empty spaces after the last input spawns (19A - 4 for product - 12 for waste - 1 for arm).

Because of this there's only two valid last moves.  Pivot the product from either fire, or slide it in the direction that only takes up 2 hexes.  Keep in mind that if we are sliding it to the output, all hexes of track must contain waste (the 2 empty spaces are on the output glyph), so the arm is completely trapped and needs to push or pull the output.  That really restricts your options, so pivoting the product consistently felt like the right choice.

I had to enumerate all layouts with the final rotation in mind to find one that would work.  Another might’ve worked, but this was the most promising and had the most degrees of freedom (being able to place the debonder and second input wherever I wanted).

In general, after the last input spawns the only way the arm can move is if one of the empty spaces is on a track hex.  Pivoting the product is great because it only requires one hex of empty space for the product to move into, meaning the other empty space can be on a track hex allowing for arm movement.

There's a few big roadblocks to a 19A that could've caused people to miss it.
1. the bonder is a pain because if there is a waste atom on one side of it, you can't place an atom on the other side without bonding them together (which usually made the situation more painful).  Once you bond the final part of the output on, you need to get one more bond on.  I solved this by prioritizing (this is by no means required for a solution to work) a set of pivots that keep the bonder entirely covered until the final pivots and drop.
2. sometimes an action required both hexes of empty space to be not on track hexes, meaning the arm was completely trapped.  Figuring out how to get the arm to trap itself while being in the right position to do the work it needed and freeing up the space it needed was a challenge, and you can see I solved that with a bonded waste pivot (cycles: who knows).

The lessons from my previous (bigger) solutions helped me here.  A similar input technique and bonded waste pivot shows up in my 20A.  That’s where I first started considering pivoting the output too, where initially I was only considering slides.

The input and debonder could be set up in another way, and the equilibrium glyph could also be moved/turned into track.  I think they’re currently in pretty fast positions though.

Reasons I think I did well on this puzzle.  I like area puzzles, but struggle with programming long tapes (helped by omclone) and with visualizing the area of large swings (not relevant to this puzzle).  I also really enjoyed this puzzle so worked hard on it.  I'm also really good at finding out how to reduce the amount of stuff I need to keep in my brain at once, which worked really well for this puzzle.  For example, I thought about  the two empty spaces instead of all the waste atoms.  There also was a lot of working backwards: start by thinking about the most constrained part of the solution (after the last input spawns).