• Summarize the findings in a google doc.
• This should include primary, secondary and sharded cluster migration cases.
• For each occurrence, what we are interested is:
  • The purpose of accessing the `config.transactions` collection.
  • How we access the `config.transactions` collection (e.g. read or write, point read or scan)

This could help us ramp up on the background, guide the design for this project as well as future projects.