Pitch accent languages are a specialized case of tone languages, ones with highly limited contrasts. The answer is the same for tone languages and non-tone languages: it depends on the details of the language. That is, each language (indeed, dialect) has a distinct intonational system.
In Eastern and Southern African Bantu languages (which generally fall in the "less tonally specified" category), the general strategy is that the pitch range expands and there is a shift upwards towards the end of an utterance, in a window of about 2-3 syllables. This alone seems to suffice to intonationally signal "is a question" (this is not how it works in English, an example of how intonational patterns are language-specific). There may be some raisings or lowering of particular syllables, but the lexical tone is generally recoverable, because H vs L is not an absolute pitch level, it is relative to pitch levels in the immediate environment, and you can tell if the pitch goes way up (probably a H) vs only a little bit (probably a L). Though, there are a number of languages with grammaticalized intonational "morphemes" marking questions, for instance final H in Shona can become L to mark a question and in principal this can lead to local neutralization between declarative sentence 1 and interrogative sentence 2. In such cases, there is also a specific yes-no question word that speakers can use, to avoid ambiguity.
There has been an enterprise of cross-linguistic description of intonational systems often subsumed under the name "ToBI", which seeks to characterize how languages differ in their intonational systems. If you are interested in a particular language, you may be able to find a basic description of the facts. See this paper, with basic references to the study.