We all made the assumption that your store_upload_data function was in an uplink program. While that technically counts as being a server call, it’s a server call that lives on the machine the uplink program is running on (probably your home machine). Every call inside that function that needs to do something with the server has to send a call out over the network to the actual Anvil server.
Every call to the actual Anvil server spins up a new instance of the Python interpreter, which delays things above what the network transit time suggests. So making a lot of calls to add_row from an uplink program is making a lot of separate server calls.
If the assumption about store_upload_data being in an uplink program isn’t right, than the comments about sending a list of dictionaries over the network don’t apply.