Are you reporting the same as The expansion of the 180-node cluster has failed by any chance? Heavy memory pressure and both in ScyllaDB 5.4. 
Is it caused by a memory allocation failure during repair?
The decoded backtrace definitely points to row_level repair codepath. You may disable it and fallback to old streaming and see if it helps. But as in the aforementioned forum post, do check how your LSA/Non-LSA consumptions were at the time of the crash and take it from there.
Same upgrade recommendations and next steps apply.