Using can in a glimmer component getter blows up dependencies about ember-can

[makepanic]

It seems like unwrapping ability.destroy() from the run call works fine and all tests pass locally.

@Exelord would you be ok with removing the wrapping run call when destroying?

from discord:

is it expected that running in a runloop creates a huge amount of dependencies for autotracking?

run() runs everything in the runloop synchronously, so if you have things scheduled that will cause updates, yes
By contrast, if you schedule into say actions normally, that will schedule it to run in the next runloop (usually, especially during render)

[Exelord]

Hey! Thanks for debugging this. I'm not in context now, why we have the run loop there. From what I remember out tests suit checks things very well. If it works without it, it probably fine to test it first in real apps :)

[makepanic]

I've already started using #110 in our app and everything is working well.