1) verbal suffix (that is, the suffix affixed to a root, dhātu, in order to make a finite verbal form, or an infinite, or a gerund)
2) verbal form (that is, the suffix+the modifications of the verbal base, except the root part)
3) finite verb
The third meaning is presupposed by classifications of words found in Nirukta and Mahābhāṣya and opposing nouns, prepositions and ākhyāta. However, in such classifications the point of what a root-part is is just not taken into account (one just classifies words as they can be actually found in usage). The first translation mostly fits, and liṅ-s are in different parts of a single text (Tantrarahasya) called both ākhyāta and pratyaya (as if the two terms were synonyms). However, 1) clashes with the above referred classification. So, the second translation seems the best one, although it presupposes a formalization independent of the actual morphemes (that is, if ākhāta were to mean “verbal form” as explained in 2), one would not be able to indicate it). A dhātu or a pratyaya can be indicated as, e.g., kṛ-, a verbal form cannot, since it “lies” both in the ending and in modifications of the root. Were Indian authors able to such a level of abstraction? Maybe yes, since they could conceive the coexistence of an ākyāta-part and a liṅ-part in the same ending of, say, yajeta. Yet, “verbal form” is not perspicuous enough.