John Chadwick wrote in the 'Case for Replacing LSJ' (BICS 11 (1994): 2):
"One of the worst features of the 9th edition was certainly not the decision of the editors at all, but of the publisher. Every user must curse many times a day the idea of saving a little space by grouping words in long paragraphs. It would matter less that each lemma does not appear at the same point in the column, if at least it were given in full in bold type. Alas, the lemma is often reduced to three letters, two or even one letter, which not even bold type can render easily visible to the searching eye. In a lecture given in Oxford in 1948 R. W. Chapman, the Secretary to the Delegates of the Clarendon Press, under whom the project was launched, had the effrontery to claim credit for this astonishing lack of perceptivity. But this passage of his lecture was wisely omitted from the published text. What other damage he may have inflicted, I do not know."
This Chapman can also be found in the app.crit. to the OCT of Plato's Laws...
The run of words ὑπάλ-ειμμα, -ει-πτος, -ειπτρον, -ειπτρίς, -είφω, -ειψις (p. 1851) frames the problem neatly. Such hyphenation follows the formation of the words (ὑπ-, ἀλειπ-) nor their syllabification (ὑ-πα-λειμ-μα, etc.).
Such printing also obscures the accentuation in the ὑπάλ-, when followed by -ειπτρίς and -είφω, gives inexplicable, unnecessary, and impossible doubling of the accent on a single word (no enclitics are involved).
If that is possible, why should ὑπαλγέω, ὑπαλεαίνω, and ὑπαλεύομαι (all hyphenless) not be grouped under ὑπαλ- also, except on the criterion of a shared root and thus meaning? Cf. the ὑο- words which share a root, but have a range of meanings depending on the second element. Better still would be to divide at ὑπ-αλγέω, -αλεαίνω, -αλειμμα, etc.
The syllabification is a puzzling facet, since no less/other than H. Stuart Jones wrote 'Appendix V: Division of Greek Words', p. 75 in Horace Hart (M.A. Printer to the University of Oxford), Rules for Compositors and Readers at the University Press, 19th edn. (5th for publication): London and Oxford, 1905.
Based on papyri of Bacchylides, Thucydides, and Hypereides, he gives three exceptions to the rule that a syllable ends in a vowel:
(1) the consonant is doubled: Συρακούσ-σας, πολ-λῷ, 'and so' Βάκ-χος, Σαπ-φώ, Ἀτ-θίς;
(2) the first consonant is a liquid or nasal (or, according to some, a <σ>): ἄμ-φακες, ἐγ-χέσπαλον, τέρ-πον, πάν-τες, ἄλ-σος; ἄν-θρωπος, ἐρ-χθέντος, ἀν-δρῶν; but βά-κτρον, κάτο-πτρον, ἐχ-θρός; θέλ-κτρον, Λαμ-πτραί.
(3) Compounds. 'For modern printing the preference must be to divide the compounds παρ-όντος, ἐφ-ῃρημένος, but ἀπέ-βη may stand as well as ἀπ-έβη.'
A different issue surfaces in αἰνό-δακρυς, ὁ, = foreg., IG12(7).115 (Amorgos) [2nd/1st c. BCE]. Note the absence of an indication of inflection (-υος cf. πολύ-δακρυς) and a date for the inscription. The foreg(oing) is αἰνο-γόνος child of praise, which is a very different meaning. Was there ever an entry for *αἰνο-δάκρυος with which 'foreg.' would have been appropriate? Cf., e.g., πολυ-δάκρυος, but note that there '= sq.' directs the reader to πολύ-δακρυς for the meaning.