Would you jam “tens of thousands” of tweets into a single Sheet / Excel file?
Nobody said Coda was scalable. On the contrary, it’s a doc and it should be treated as if it were a doc, not an ever-expanding database. It’s easy to forget behind the “it’s in cloud and everything magically works” argument that in the end all is tech and tech has limitations.
If you want to get tens of thousands of tweets into a single doc, then probably you should reconsider the architecture altogether. Do you need to actively calculate on all of the tens of thousands with real time formulas? Or will it only be a portion of data that a person interacts with? You say you run 500 marketing campaigns — IMO in your context it makes sense to have one doc per campaign or one doc per client, one master doc for overall tracking, and cross-doc connections in place (e.g. with my upcoming Advanced Doc Connections pack)