In many stable bilingual populations--including in Canada's capital region--code-switching is common in-group behaviour. Though code-switching is described by some as "the most comfortable way" to speak (Dorleijn, 2017: 16), empirical research continues to associate even highly habitual code-switching to behavioural and neurophysiological indices of effortful processing (e.g., Gosselin & Sabourin, 2021). In parallel, self-reported code-switching habits have been linked to increased inhibitory control abilities (e.g., Gosselin & Sabourin, 2023), suggesting that switching is indeed cognitively demanding. When is this 'code-switching cost' found, and what does it indicate? Is it reasonable to compare real-life language production to in-lab artificial processing? This talk details the ways in which my dissertation attempts to disentangle the production-processing paradox in the code-switching literature. In brief, I will introduce FEBLOC (French-English Bilingual Loved Ones Corpus), a novel corpus constructed with the purpose of examining the topic through a naturalistic lens. I will present preliminary data supporting the uncoupling of bilingual language production and bilingual language processing.