From the Hampster Dance, All your Base are belong to us and the Dancing Baby in the second half of the 1990s to Bernie’s mittens at the US presidential inauguration and the image macros of the Evergreen blocked in the Suez Canal, through Disaster Girl or Distracted Boyfriend to name but a few, memes and Internet phenomena have become in the last twenty years an important part of our digital cultures (Shifman, 2014). After presenting why historicising virality matters (which is the purpose of our Hivi research project supported by the Luxembourg National Research Fund and conducted at the University of Luxembourg, https://hivi.uni.lu), this talk will focus on the challenges related to sources. Historicising online virality requires to first acknowledge how difficult it is to build corpora and develop chronological views in the vast and heterogeneous amount of data lakes and sources available on the live web and in web archives. All these repositories preserve viral content in a very different manner, covering different frames and periods, while the actors who come into play have different size and motivations, from giant platforms (Twitter, YouTube - see Burgess and Green 2018, etc.) to small communities, through heritagisation platform (i.e. Know Your Meme - see Pettis, 2021) and web archives. Therefore, it is necessary to consider the many levels of politics, curation and agencies at stake when studying the history and heritagisation of memes and Internet phenomena and their impact. Finally, temporalities and scales are also challenging, and scalability is key for studying the spreadability of digital contents, notably in a long-term perspective. There is a constant need for balancing between several scales to study circulation and flow (Jenkins, 2009), processes, participation (Milner, 2018) and appropriation, while recontextualizing memes and Internet phenomena in their complex and changing environments (changes of platforms, of participation, of audience, of meaning...).