What most zap providers don't realize is they have to be careful with state management.

Unlike receiving LN, receiving zaps = active tracking of pending invoices = memory + ongoing CPU cycles

Unless a zap provider adds a sort of "max pending zaps" ceiling, above which it wont accept new ones, until some are processed or expire -- then they're like sitting ducks for script kiddies.

That's the ultimate stress test IMO. On loop, open as many pending zaps as possible (but don't pay).