Currently, we rely on various objects in scraperlib to:

• create the ZIM
• re-encode videos and images
• cache these assets on the optimization cache

We might consider to have a mechanism to perform sanity checks on scraper behavior:

• did we cached all re-encoded images / videos when a cache is present?
• did we removed temporary files from the filesystem as they are added to the ZIM? (we know that while we prefer in-memory/streaming approaches, there are still many scrapers which are using the temporary file approach, and even some situation which have to rely on it)

What I do not yet know:

• should we make the scraper fails if these checks fails?
• is there any chance we automate these checks? (i.e. no need to modify the scrapers, or as little as possible - at least not make a call to "check_i_m_ok" mandatory, because the scraper developers might forget about it as well ; I doubt about this because there are many kind of situations)
• can we do these checks early? (so that we fail the scraper asap instead of wasting time and resources)