The first thing about indirect questions is that they are not questions, either in English or in toki pona. Two things make them feel like questions: 1) interrogative pronouns look exactly like relative pronouns in English (and most I-E languages since PIE) and 2) the answer to a question is often (among other things) the corresponding “indirect” form: “What did he throw?” is answered tautologously with “What he threw”(if the interrogative word is the subject, the forms are even identical in sound-- barring intonation --: “Who threw it”). This comparison shows the difference in English: questions are pronoun tense subject verb oblique, relatives are pronoun subject verb oblique. Questions are sentences, though twisted ones usually; relatives are nouns or adjectives.
This relation gives rise to the usual explanation (other than mere calquing) for using toki pona question for for “indirect questions”: they are short for “the answer to the question given”. While this explanation has, as we will see, some advantages, it hard to justify even as a calque and certainly does not fit into toki pona grammar. At its best it is akin to confusing a statement that one has a desire with an expression of that desire (or maybe directive for that desire to be fulfilled). The place where an “indirect question” goes calls for a noun phrase and, in toki pona, sentences are not noun phrases (they are in English, but, as noted, “indirect questions” are not even sentences).
I am going to assume that most relative clauses are not a problem, that no one would translate “The man who came to dinner stayed a month” a ‘jan seme li kama tawa moku? ona li awen lon tenpo mun’ nor “I want what you have” as ‘mi wile e ni: sina jo e seme?”. The problem cases seem to center on ‘sona’ and occasional other words used with related meanings. This is somewhat ironic, since ‘sona’, in the appropriate sense, requires its object to be something that is either true or false (a proposition, in short) and question definitely are not (they may be both or neither or something else entirely depending on how you work it). So ‘mi sona e ni: sina tawa seme?’ is very odd for “I know where you are going”, in ways that either “mi sona e ni: sina tawa ona’ or ‘mi sona e tawa sina’, the two normal “relative clause” versions, are not.
All of these, even the “indirect question” form run into problems with ‘sona’. toki pona follows English and not the major European languages in having a single “know”, rather than separate “know facts” (wissen, savoir, saber) and “be familiar with people and things” (kennen, connotre, conocer) (This is the opposite of the the “make/do” division where it follows Europe in not having the distinction -- or shuffling its use). So we have ‘mi kama sona e toki pona’ “I learn toki pona’ but also ‘mi kama sona e jan Mali’ “I meet/ get acquainted with Mary”. The upshot of this is that, given ‘mi sona e tawa sina’ or any of the variants, we are not sure whether I am familiar with the place -- which I might be because I know it is one of three places I know well -- or I know the identity of the place you are going to, i.e., that you are going there. We have the same problem in English, of course, but seem to avoid its consequences for the most part, though I can’t figure just how. In the cases of mysteries, we often have the situation where everyone knows the murderer, the person who, it will turn out, is the murderer, but (almost) no one knows who the murderer is. This looks like inserting a separate notion at some level in the development of our sentences “I know the person who is the murderer” vs. “I know the identity of the murderer”. We can do a similar thing in toki pona. Or we can leave toki pona as it is and hope that context will resolve the issue.
Notice that using “indirect questions” doesn’t help with this issue at all, except that we can just ask different question ‘seme li moli e jan?’ (or ‘jan pi moli jan li seme?) and ‘semi li sama pi jan pi moli jan?’ (and various variations). There is one indirect question where the question form might help. Oddly, it is a form that doesn’t have a direct question easy parallel, “whether”. “Whether you are coming?” or the expected “Whether are you coming?” isn’t an English sentence (and hasn’t been for at least a long time). The corresponding question is actually “Are you coming?” and there is no interrogative pronoun for this -- nor a ‘seme’ in the toki pona pattern ‘sina kama ala kama.’ So, how do we say “You know whether they are coming” in toki pona? Before we press ‘sina sona e ni: ona li kama ala kama?’ into service, we might try the ‘sama’ trick of ‘sona’ problems and say “You know the truth value of that they are coming” ‘sina sona e lon ni: ona li kama’ or ‘sina sona e lon pi kama ona’ But that looks mighty like “You know the truth of that they are coming”, i.e., “You know that they are coming”, so we need at least a better word for “truth value” (or whatever) to fill the gap and this is getting rather sophisticated for a simple language (as if “truth value” and “identity” weren’t bad enough).
But, if we allow Y/N questions in as objects of at least ‘sona’ in toki pona, what grounds to do we have for excluding WH questions from the same role? Isn’t this just caving to English and not considering how things work autonomously in toki pona? But, as noted, English doesn’t use questions as objects of “know”, despite what various people (including some scholars) say. toki pona might have autonomously taken this path which English only appeared to approximate. (This line of chat seems the worst sort of rationalization, except that it works so well.) Further, the use of questions dodges the ‘sona’ problem. ‘mi sona e tawa sina’ or ‘e ni: sina tawa ona’ is simply a different claim from ‘mi sona e ni: sina tawa seme?’ with just the right difference.
When we move away from ‘sona’ to other cognitive verbs, however, we run into the problem that most of these (‘toki’, ‘kute’, ‘sitelen’, ‘lukin’, ‘pilin’) can take not only indirect question kinds of objects but also the questions themselves as quoted objects: “He said who the murderer is” “He said ‘Who is the murderer?”, both now presentable as ‘ona li toki e ni : jan li moli jan li seme?’ We can put the question in quotes, but are not obliged to. We could require this, but that has not been a favored response to use/mention problems in tp. So, we are left with problem we started with, though in a different configuration. Ultimately, whatever we do, we have to rely on context to get us to what is meant. And good luck with that.