Chrome/FF seems to parse :before#id or :before.class as invalid (instead of valid and matching nothing) but I do not see on which part of the spec this is based.

Pseudo-elements are featureless, and so can’t be matched by any other selector

This does not make them invalid.

Legacy : pseudo-element syntax is defined in <subclass-selector> as a <pseudo-class-selector>.

<compound-selector> = [ <type-selector>? <subclass-selector>* [ <pseudo-element-selector> <pseudo-class-selector>* ]* ]!

Subclass selectors can appear in any order while still selecting the same set of elements, but this is not true for a pseudo-element specified with the legacy single : syntax.

Therefore I assume that :before must be handled as if it has been specified with ::before and matched <pseudo-element-selector>. If so, it may be helpfull to define it in the spec.

A similar confusion occurred to me when parsing a <color-stop-list>, whose syntax requires two color stops but its grammar accepts an implicit color stop defining two stops with the same color.