We check the stability of the ABMP16 fit with respect to
modifications of quark PDFs suggested in the recent literature: the strange sea enhancement and a positive non-vanishing $d/u$ ratio at $x \to 1$. These possibilities are examined using test versions of the ABMP16 PDF fit which demonstrate no need of those changes. Furthermore, we localize peculiar features in other analyses which are responsible for a different behaviour of the PDFs obtained.
The strange sea enhancement can be explained by a choice of the
PDF shapes being not flexible enough and therefore leading to an over-suppressed $d$-quark distribution. This suppression has to be compensated by a corresponding rise by the $s$-quark distribution.
As a result, an unusually large strange sea suppression factor
is obtained. The non-vanishing value of $d/u\vert _{x=1}$ becomes consistent with zero in case the higher-order QCD corrections to the D{\O} $W$-asymmetry data, which drive its value, are taken into account. Furthermore, the related $e$-asymmetry sample prefers a slightly negative value of $d/u\vert _{x=1}$, although it is consistent with zero. These clarifications support confidence in the PDF shapes used in the ABMP16 analyses.