This solve is vaguely inspired by madmaster's icelandic lava salt mobius conveyor.  I wanted to build a polymer which eventually passed through the center at a column every cycle, and which evolved over time using a feedback loop.

However, I'm addicted to chaos.  And as such, I decided to try to engineer a nearly chaotic feedback loop, such that the eventual success took many loops to reach.  So here's how it works:

The top row of the central polymer is always earth, nothing can modify that.  The bottom row varies over the course of the feedback loop, which panic helpfully verified was 107280, a healthy six digit number.

When the left atoms aren't arriving, the polymer moves at half speed and retains a bottom layer of salt which gets turned into elements by the spinny Berlo.  The timing for this case ends up eventually feeding these pairs back through the polymer from the left.

When those atoms are being fed back, the polymer moves at full speed and the lower row has doublets of everything that was fed back from below.  So the win condition is effectively 6 or more earth in a row on the bottom.  But at least at first, all of the atoms we see are air, water, and fire.

Once the full speed polymer gets to the right side, the mors that's been made all along manages to wand it out of the way, hitching a ride all the way around the entire feedback loop.  At the top, it gets partially fed back for input suppression, and partially checked for quintessence.

The somewhat chaotic nature of the quintessence procs means that it mostly fails, but it is capable of succeeding.  Once it does succeed, that quint has to make a daring escape including pog swings through 3 different rotations of the mors.  After that, it enables an "off" switch on the high speed portion of the polymer feedback loop.  That switch remains off, diverting all chunks to a lower right waste chain, until the effect propagates all the way around and makes the polymer go back to half speed.  At this point, the smaller chunks make it past the off switch, and the loop switches back on.

These delays lead to corresponding delays in making quintessence, which controls the actual timing of when the switch goes on and off.  So the flow of atoms around the loop is super hard to predict.  Berlo keeps on spinning and eventually some earth shows up in the loop.  It can be doubled, but if the polymer is moving at full speed then almost half of the atoms get sent to quint instead, so the doubling is ineffective.  It relies on some combination of delays to build up a large chunk of contiguous earth.  This happens on cycle 2300ish.

I love this solution so much.