This task is also re: thanos-swift capacity planning, due to thanos itself storing more data and other projects also adding data (notably tegola/maps), the space on thanos-swift is growing at about ~13% over three months, and now at ~80%, meaning space will run out in ~a quarter.
Also related to thanos-swift and MOSS, there will be two hosts moved from thanos-swift to MOSS purposes to implement ceph, and thanos-swift replication factor will be moved from 4x to 3x.
To ease all the operations above I propose we move from current 270w retention (~5y) to 110w (~2y). The oldest data we have in Thanos is a little over 2y old (June 2020) so this will effectively "cap" the data at more or less the current usage (modulo organic growth)