On demiexplosive rules

[Hdjensofjfnen]

As a truth of non-totalistic cellular automata, some are explosive.
Others are not explosive.
But there is a small group of rules that are demiexplosive - such that asymmetric patterns stabilize, but symmetric patterns explode. (As in the old joke, C4 in these rules will explode.) Why is this the case?

[Layz Boi]

Why is this the case?

I'm no quantum rocket surgeon, but I want to guess that it's because they have a partial of ship/breeder/stretcher/moving-object that easily occurs in a particular symmetry. This theoretical partial also should have the same symmetry as the patterns' symmetry along the edge of the soup it spawned from, while not being extremely simple. This then readily grows the pattern primarily along one axis.
Or something of this nature. I dunno for sure.

[testitemqlstudop]

However there are many more demiexplosive D2_1 rules than D2_2.

[praosylen]

I personally find rules that are stable on C1, not-quite-explosive on D2_1 or D2_2 in addition to D2_x, and extremely persistent on D8 quite fascinating -- some of the current methuselah record holders are in these rules, which tend to maintain a central active region, send out symmetric tendrils periodically in all directions leaving various debris, and after a huge number of generations stabilize in the center through sheet random chance, reminiscent of something from (e.g.) ExtendedLife.

[77topaz]

Rules with sufficiently complex Margolus mechanisms in them like Wild Seas seem to tend to fall in the D2_1 demiexplosive category.