A tx size limitation that is weakly enforced would be a start, eventually the lack of support for big ones would amplify the size cost of work-arounds on newer branches could be closed up once a minority of miners are running them. It is possible to discover which nodes are originating these blocks containing inscriptions.

When I say "amplify the size" i mean by forcing multiple transaction overhead on the transactions to chunk the inscriptions out.

A gentle approach could work, and it would steadily increase the latency of propagation of competing blocks that would otherwise reorg, to favor the ones with no or less big transactions.