It just seems like front-ends aren't equipped for automated tantrums. It's pretty informative that someone can perform what amounts to a denial of service attack with relatively little stake at their disposal.
Maybe we need to go in the direction of front-ends fully ignoring tantrum throwers, which means not accessing content directly from steemd anymore and instead making heavy use of consensus interpreters (like hivemind, only more tuned into what tantrums look like so they just don't index them).