I don’t understand what you mean. Every full node always processes all blocks content, whether it is right after the block was created (what i believe you call “at tip”, am i correct?), when it catches up after some downtime, or when it performs IBD.
Adding a new data structure to consensus rules means every single full node will have to download and validate this data. (In this case as i understand it, download the SegData and verify the Merkle root commitment.)
Furthermore, this additional data needs to be served by some nodes. If most reachable nodes do not serve it, this would impose significant load and reliance on the few that do.
I am unconvinced that you can simply assume so, but let’s take it for granted. This is still a block size limit increase, and because there is unlimited demand for free replicated storage, i think we can expect that space to be filled if it’s made available.
That comes back to my previous point, but no: every node that wants to fully validate the state of the system will have to process any additional data structure introduced.