recently i've been thinking about lojban morphology and how it somehow works, despite being seemingly put together with scotch tape and tears from failed dreams.
if you've ever looked into lojban morphology for a while, you know that there's this thing where some words, if they aren't compounded correctly, will «fall off». the usual example is *tosmabru, but that isn't even a real word that means anything, so let's use *bisladru instead ('ice cream').
-bis- is the combining form of bisli 'ice' and ladru means 'milk'. but you can't just put them together like that, because ×bisladru sounds like bi sladru (and that would mean 'eight celebratory roofs', with the bi being a particle, and the sladru another compound word)
so the actual form has to be bisyladru, with a -y- keeping the word from falling apart.
this is problematic when you get to another kind of word though — so-called slinku'is. i won't really consider zi'evla in this analysis (i think i could if i really tried to, but it would be very hard), so the question i'm going to try to answer here is «why does bavlamdei 'tomorrow' or jostolcru 'to ban' not fall off?»
looking at those words, and knowing that you have to be careful so that words don't «fall off» in lojban, you'd imagine that bavlamdei would parse as ×ba vlamdei, and jostolcru → ×jo stolcru. but that is not what happens, and those two words are completely valid, while *vlamdei isn't even a valid lojban word. why?
my (theoretical) answer to this is that it has to do with syllable weight. lojban syllable structure is, at most, CCCVC (the maximal example is the first syllable of strimgau 'to stream', which is a zi'evla, and i'm not considering those). but, as far as only the «native» vocabulary of cmavo, gismu, and (non-extended) lujvo is concerned, the most complicated syllable format allowed is either CCV, CVC, or CVV, but you never get combinations of these, like CCVC or CVVC. (the only exceptions to this involve epenthetic r/n in lujvo: fairgau 'to share' from fai-gau.)
this actually allows for a «cute» formalization of the tosmabru and slinku'i tests: the first syllable of a valid word, if it is unstressed, must be closed. bemjoitco 'of the americas', darvistci 'telescope', kertinytci 'headphones', all pass this test, but so do prulamdei 'yesterday', gliglibau 'england english' and brivlacme 'word name' (as opposed to a name name, lol), because in lojban, it makes the most sense to consider a CCV syllable to be «closed»...
(on the other hand, this means that ×tosmabru ×bisladru ×sozmasti are not real words, because those can be split as to-sma-bru, bi-sla-dru, so-zma-sti, and the first syllable is open. bisyladru and sozymasti are fine, though, if we only posit one rule: y forms a syllable of its own, and won't take a consonant into its syllable unless it has to. so bisyladru is bis-y-la-dru.)
now to slinku'i-ish cases, like bavlamdei. this seems like it would fall off under our new rule, because it's ba-vlam-dei, but it really isn't. the «correct» syllabification is bav-lam-dei. remember what i said about CCVC like -vlam- not being a valid native syllable except for the special cases of inserting epenthetic consonants? well, this here means that a syllabification like bav-lam-dei makes more sense than ba-vlam-dei, because it avoids a single syllable with 3 consonants in it (which as we've seen does not happen outside of exceptional cases), and so bavlamdei passes the test; the first syllable is indeed closed.
okay this might have made zero sense to you but i hope you liked it if you were at least able to understand it 
ask me if you have any questions please... ki'e sai tcidu mu'o iu