[seqfan] Need proof that Kimberling's A26185 = A198173

[Neil Sloane]

Dear Seq Fans, Michel Dekking (<a
href="https://arxiv.org/abs/2001.08915">Permutations
of N generated by left-right filling algorithms</a>, arXiv:2001.08915
[math.CO], 2020) recently showed that many of Clark's sequences which
appeared to be duplicates really were duplicates.

I have now processed all these pairs or triples or even quadruples of
identical-valued sequences and declared  a dozen or more of these entries
to be duplicates with keyword "dead".

But one of these triples has a loose end. A026209 = A026185 is true,
but  A026185 =? A198473 is an open question.  The latter has a simple
definition:
If n even, then 2n. If n odd, then nearest integer to 2n/3.

but A026185 is not so easy to describe.
Could someone find a proof?  Then we can merge the entries and declare one
more sequence to be "dead".