Some MultiPDepArrhenius rates in libraries don't validate in Cantera

The NOx2018 library index = 258, label = "C2H3 + CH2O <=> C2H4 + HCO", has a MultiPDepArrhenius rate expression that comes from (and generates) DUPLICATE sets of PLOG rates in Chemkin format. When translated to Cantera, it chokes with

***********************************************************************
CanteraError thrown by Plog::validate:
Invalid rate coefficient for reaction 'C2H3 + CH2O <=> C2H4 + HCO'
at P = 101.325, T = 200
***********************************************************************

The automatic validation checks an array of T and P, and reveals that at 200K the net rate (the sum of the two expressions) is negative.

The source of the NOx2018 library is
Peter Glarborg, James A. Miller, Branko Ruscic, Stephen J. Klippenstein, Modeling nitrogen chemistry in combustion, Progress in Energy and Combustion Science, 67, 31-68, (2018)
https://doi.org/10.1016/j.pecs.2018.01.002.

The chemkin file contains a comment pointing to a @cfgoldsmith paper on the second (negative) term. This term is exactly the same expression as the following reaction which has the same reactants but 3 products C2H4+H+CO. I guess the trick is to subtract the "hot" route direct to H+CO from the HCO channel and add in the stand-alone reaction. Makes sense. But at low temperatures, this expression is faster than the original, and the net rate ends up negative.

The original they used appears in several other models, including AramcoMech 2, though I'm not sure who used it first.

C2H3+CH2O=C2H4+HCO                   1.4E10   0.209    3934 ! 1 atm
 PLOG/ 0.001                         1.1E07   1.090    1807/
 PLOG/ 0.01                          2.5E07   0.993    1995/
 PLOG/ 0.1                           2.5E08   0.704    2596/
 PLOG/ 1                             1.4E10   0.209    3934/
 PLOG/ 10                            3.5E13  -0.726    6944/
 PLOG/ 100                           3.3E14  -0.866   10966/
 PLOG/ 1000                          1.7E01   3.170    9400/
 DUPLICATE
C2H3+CH2O=C2H4+HCO                  -8.5E19  -2.264   23862 ! 1 atm
 PLOG/ 1.000E-03                    -2.3E16  -1.269   20617 /
 PLOG/ 1.000E-02                    -5.2E16  -1.366   20805 /
 PLOG/ 1.000E-01                    -1.5E18  -1.769   22524 /
 PLOG/ 1.000E00                     -8.5E19  -2.264   23862 /
 PLOG/ 1.000E01                     -4.4E23  -3.278   27795 /
 PLOG/ 1.000E02                     -4.2E24  -3.418   31817 /
 PLOG/ 1.000E03                     -2.1E11   0.618   30251 /
 DUPLICATE
! C.F. Goldsmith et al., J. Phys. Chem. A 116 (2012) 3325�3346.
!
C2H3+CH2O=C2H4+H+CO                  8.5E19  -2.264   23862 ! 1 atm
 PLOG/ 1.000E-03                     2.3E16  -1.269   20617 /
 PLOG/ 1.000E-02                     5.2E16  -1.366   20805 /
 PLOG/ 1.000E-01                     1.5E18  -1.769   22524 /
 PLOG/ 1.000E00                      8.5E19  -2.264   23862 /
 PLOG/ 1.000E01                      4.4E23  -3.278   27795 /
 PLOG/ 1.000E02                      4.2E24  -3.418   31817 /
 PLOG/ 1.000E03                      2.1E11   0.618   30251 /

Any suggestions, @cfgoldsmith? I can think of...

1. just revert to the AramcoMech rate and not use the hot route.
2. change Cantera to not do these sanity checks? (or not do them at 200K)
3. ...?

Start by reproducing the NOx2018 library index 258 translation and Cantera validation at 200 K and 101.325 kPa. Inspect the duplicate PLOG rate sets and the related C2H3 + CH2O = C2H4 + H + CO reaction described in the issue. Done means an agreed correction or validation policy is implemented and the translated rates no longer fail validation.