Hypothetical main size determination in flex layout algorithm  has unnecessary assymetry wrt flex items parallel vs. orthogonal to main axis. The distinction is not useful, because (1) items in parallel writing mode can have orthogonal content and (2) there are other reasons for item layout to behave similar to orthogonal writing mode, e.g. a multiline flexbox with orthogonal main axis.
What matters is dependency of content max-size an min-size on cross size of flex container.
The two bullet items on laying out items in available space should be replaced with following:
Otherwise, lay out the item using available space with following dimensions:
* On main axis:
- if flex-basis is auto: infinite
- if flex-basis ‘fit-available’ or ‘fit-content’:
flex-container’s main size, constrained by its
min and max size.
* On cross axis: flex container’s cross size, constrained
by its min and max size.
The flex base size is the item's resulting measure.