I found that it's possible to have three different versions of la. Right now, pretending dotside is official, we have two versions: cmevla- and brivla-la, the former requiring denpa bu before and after the cmevla, the latter requiring none.
In the spirit of the above texts, however, it seems natural to keep the pre-dotside version of la alive along with the other two versions.
The advantage of this is that denpa bu can be elided more often. More specifically, any cmevla that does not contain "la'', "lai'', or "doi'' would not need a denpa bu before it. Conversely, if you hear {la.}+Consonant, you know right away that the cmevla might contain la/lai/doi.
The obvious counter-argument to keeping this "old la" is that it's "too difficult" to always know ahead of time what letters a name is going to be containing and thus people will make an error sooner or later. This argument, however, can be refuted by the fact that nobody is forced to use this old version of la and instead is able to, when in doubt, use the dotside version. The advantage becomes clear once you *do* know for certain what name you're going to utter, because then you can elide denpa bu like in old times.
Combining all the things mentioned so far, we can finnally elide all the denpa bu in the following bridi:
(1) .i la bab klama
"Bob goes." — because bab contains no la/lai/doi, we can use old-la {la}, and since *{bk} is forbidden, bab klama falls apart as bab klama (and not {babklama}.

For reference, here is the usage restriction for old-la:
Names are not permitted to have the sequences ``la'', ``lai'', or ``doi'' embedded in them, unless the sequence is immediately preceded by a consonant.

(This means that a name like {klark} could be used with old-la.)

This concludes my musing on denpa bu, cmevla and la.

It is possible to elide denpa bu in several phonological contexts, which is generally positive but which has the downside of being a bit irregular. Here are some situations where denpa bu is elidible.The denpa bu after lerfu of selma'o BY.
This one is sometimes elidible and sometimes not. It's a dangerous area because it can really screw with your bridi. Consider:
{nu ny. nanmu} vs {nu ny nanmu}
"an event of N being a man" vs "an event of being a man" (it parses as {nunynanmu})
So in this case, eliding denpa bu after BY will lead to a catastrophy.
We can generalize: Whenever we have a string of letters of the form *CV.CyC* (where the second and third C are identical), we can't elide denpa bu. (the * represents any string of letters).

Also note that the above problem can always be avoided by using {cu} between the lerfu and the selbri, which again allows us to elide denpa bu.

Denpa bu before the approximants [w] and [j] ({uV} and {iV} respectively)
As it turns out, denpa can often be elided before words beginning in {uV} or {iV} because these aren't really vowels in those contexts. They are approximants and are more like consonants than vowels in Lojban.
In {.i ui mi klama}, {ui} doesn't need the denpa bu because [w] is not a vowel and it parses automatically.
When pronounced clearly {.i ia} becomes [ʔija], (which has both the vowel [i] and the approximant [j] in it, while the word {ia} should never have an [i] in it.) and as such is not ambiguous.

I seem to be the only one to use and understand denpa bu this way.
First of all, I hear people speak of "pauses" and saying things like "I wonder if I make long enough pauses..". This is completely alien to me. Linguistically speaking, denpa bu is /ʔ/ which should normally be realized as [ʔ] and has the possible allophone [ ] (no sound) Yet, people seem to think it's just a "pause" while forgetting about [ʔ] in most phonological contexts. And while this is of course technically legal, it is also the cause of many misunderstandings and causes wrong usage / incorrect definitions of things like zo'oi (which is already broken enough).
Inconsistent usage of denpa bu makes audio-visual isomorphism (AVIM) fall apart. Therefore, people need to start accepting it as a normal letter that does not differ from any of the other letters (e.g. {b} or {t} etc). It might look different, but it's an equally important part of the language.
{.i} should be pronounced as [ʔi] and nothing else.

Now about dotside. It's a good proposal and certainly better than having broken LA.

To me, and this might surprise you, denpa bu is the delimeter of LA. Unlike the ZOI delimiters, it's an enclitic, for the simple reason that anything else would be unnatural. Consider the name {la pant}. Most people either forget the dots entirely, or they place them thus: {la .pant.} which just shows they don't really know what to do with denpa bu. Lojban has AVIM and thus writing it differently from how you would actually say it is wrong. Now it *is* possible to pronounce this the way it's written, namely [?pant?], but I don't think most people do that. Are they just waiting to let their so-called "pause" go by before they pronounce the word? And then another "pause" right after it? I hope they don't. A much more natural way to pronounce it is to move the glottal stop to the end of the article: {la.} [laʔ]. This is quite normal in many languages and easily manageable for anyone else. The next denpa bu we can move around freely, since it's practically just the terminator of the first one. If you are able to pronounce pant as [pantʔ], go ahead and write {pant.}. Some can't pronounce that and therefore shouldn't write it that way. In many cases, the second denpa bu can actually be elided, because the next word starts with one. If we had for example *{la pant e mi}, the .e would let us elide the terminating denpa bu yielding {la. kant .e mi} which is perfectly fine. If the following word didn't start with denpa bu, we'd have a different situation. Say we have *{la pant sipna}. Here we have three possibilities:
    (1) {la. pant. klama}
    (2) {la. pant .klama}
    (3) {la. pant . klama}
The first I already talked about. The second is quite feasible, since it's similar to a glottalized k in that case, which exists in many English dialects. This might be more problematic for things like ?{la. pant .barda}, but it's possible. The third option is the most flexible and not dependant on neighboring words. It also shows how denpa bu functions as a terminator. To the untrained ear it will sound like a pause, but it's not. You are saying [la? pant ? klama]. Personally, I think both (2) and (3) are accaptable.

Slight addition: For cmevla that begin in a vowel (this does not include approximants), the "pre-dotside {la}" can be used. That is, you don't say {la. aipod}, but {la .aipod}. Why? Because no word can begin in a vowel. *{aipod} would break this law. You should think of this word as beginning in a consonant (.), but since this consonant is {.}, you'd end up with {la. .aipod}. So in this case, you can omit the first of these denpa bu. Also note that {la .aipod} can not fall apart as {la .ai + pod}, because for that there would have to be another denpa bu before {pod} like so: {la .ai. pod}. As you can see, the rules for  denpa, once you understand them, are completely stable and unambiguous.

Now on to ZOhOI. The definition of it is horrible. Here it is: "quote next non-Lojban word only; quotes a single non-Lojban word delimited by pauses (in speech) or whitespace (in writing)". Do you notice anything? It totally breaks AVIM because it has different rules for speaking and writing which is as anti-Lojban as it gets. So if you want to use it, apart from it being broken for other reasons as well, you *must* use denpa bu as delimiter. *{la'oi George} has to be {la'oi. George .} for it to be correct. (This will also allow you to say {la'oi. George Washington .} because there isn't actually a pause between the first and last name; they are a single word. Writing it this way will fix a lot of ambiguity.)

One last bit: At the end of text as well as at the end of a bridi denpa bu can be elided, the former because there's nothing you'd need the cmevla to be seperated from, the latter because a following {.i} already has a denpa bu.
I often see people writing {mi viska la .pant. .i co'e}, which is unnecessary and most likely doesn't correspond to how they pronounce it either, unless they say [pantʔʔi], which I doubt.

As far as I know, I'm the only one who uses denpa bu in the way recommended in this article, but it's the only way that makes it pronouncable in real life and that doesn't violate AVIM. Please consider this with an open heart.