When resharding recipient is running under wt cache pressure. It's operation can get killed by WT and throw an error that is currently not retryable:

shard-2 reached an unrecoverable error :: caused by :: WiredTigerIdIndex::insert: index: _id; uri: table:index-81229-14547801253578948280 -31800: oldest pinned transaction ID rolled back for eviction

I believe this is normally handled in other write paths with the write conflict retry loop. We should investigate whether resharding can safely retry this.