mypy struggles to prove this cast correct when there are two or'd
isinstance checks where both types coincidentally have properties of the
same name (but no defined protocol making that explicit). I'm not really
sure why mypy is happy with this in its current state, but it isn't
after a change I'm making.
All our isinstance use is a bit of an anti-pattern anyway, so extract a
method that exposes the data we care about.
The start/end times for tankers aren't actually used, so this could be
simplified even more, but that data _should_ be used.