Does it make sense to merge these two together? Something like

const suggestions = React.useMemo(() => {
  if (!typeaheadMatchedStrings) {
    return [];
  }
  const suggestedUsers = getMentionTypeaheadUserSuggestions(
    userSearchIndex,
    props.inputState.typeaheadState.frozenUserMentionsCandidates,
    viewerID,
    typeaheadMatchedStrings.textPrefix,
  );

  const suggestedChats = getMentionTypeaheadChatSuggestions(
    chatMentionSearchIndex,
    props.inputState.typeaheadState.frozenChatMentionsCandidates,
    typeaheadMatchedStrings.textPrefix,
  );

  return [...suggestedUsers, ...suggestedChats];
}, [
  typeaheadMatchedStrings,
  userSearchIndex,
  props.inputState.typeaheadState.frozenUserMentionsCandidates,
  viewerID,
  chatMentionSearchIndex,
]);

Yes, my motivation to split this code into three useMemos was that if userSearchIndex changes and chatMentionSearchIndex not, then we would recalculate userSuggestions only, but the only time userSearchIndex and chatMentionSearchIndex would change is when user switches the focused chat, so my intuition was def wrong.

Why one of them is an array and one is an object?

It might make sense to extract this as a function to a more discoverable place.

frozenChatMentionsCandidates is an object, because it's faster to get thread info from threadID in getMentionTypeaheadChatSuggestions. frozenUserMentionsCandidates is an array, because we handle user mentions differently in getMentionTypeaheadUserSuggestions