Characters missing from R-generated graphic when \pdfinclusioncopyfonts=0

I have a .pdf figure generated by R/ggplot2, which renders correctly when viewed as a standalone file, but when embedded in a document by pdftex, one character is missing. I've cut the problem down to this R script...

library(grid)
cairo_pdf("arghfig.pdf", width=2, height=0.5)
grid.newpage()
grid.text(label = "Bias (\U01d712\u00b2 odds ratio)") # 𝜒²
invisible(dev.off())


plus this wrapper LaTeX file:

\documentclass{article}
\usepackage{graphicx}
\begin{document}
\pagestyle{empty}
\includegraphics{arghfig.pdf}
\end{document}


The actual PDFs produced by this are available from https://hacks.owlfolio.org/scratchpad/font-inclusion-bug-2016-03-18/. (For your convenience, the -q variants of the PDFs have been internally decompressed using QPDF.) Here's what the standalone file looks like...

and here's what happens to it when passed through pdflatex...

\pdfinclusioncopyfonts=1 is a workaround, but I have a lot of text-heavy figures in this document and it doesn't appear that one can activate this mode for just one figure, so I'd prefer not to have to do that. Also I would like to understand why only the superscript 2 is affected, when it's my understanding that \pdfinclusioncopyfonts problems should affect all characters that aren't used in the surrounding document's text, and this wrapper document has no text.

Useful answers to this question would include any or all of:

• an explanation of exactly what has gone wrong here
• an informed opinion as to whether this is a bug in R core, grid, cairo, pdftex, or what
• suggested workarounds in the R script (note: grid is mandatory, because ggplot2, and plotmath is forbidden unless you know a way to do multi-line plotmath captions with sensible amounts of line-spacing, i.e. not atop())
• suggested workarounds in pdftex other than \pdfinclusioncopyfonts=1 (note: luatex is not an option, because egregious and unmodifiable journal style file).
• suggested workarounds via hand-hacking the PDF figure (not kidding! I have to do that anyway)

Note that there are two embedded fonts in both documents...

$pdffonts arghfig.pdf name type encoding emb sub uni object ID ------------------------------- ------------- ------------ --- --- --- --------- HLICDJ+TeXGyreHeros-Regular Type 1C WinAnsi yes yes yes 5 0 JMSOLB+FreeSerif CID Type 0C Identity-H yes yes yes 6 0$ pdffonts argh.pdf
name                            type          encoding     emb sub uni object ID
------------------------------- ------------- ------------ --- --- --- ---------
ZJEQCD+TeXGyreHeros-Regular     Type 1        Custom       yes yes yes     11  0
JMSOLB+FreeSerif                CID Type 0C   Identity-H   yes yes yes     12  0


... but the superscript 2 is coming from the Type 1(C) font, not the wacky "CID Type 0C" thing. That font only provides the 𝜒 character.

 d:/texlive/2015/texmf-dist/fonts/type1/public/tex-gyre/qhvr.pfb

\pdfinclusioncopyfonts=1 forces pdflatex to use the font in graphic instead of searching a replacement, imho this is the right thing to do if you can't change the way the graphic embeds and uses the font.