In a simplified system suppose that we just double easy card review times, and reset hard cards back to 1. You have a card up for review today, you last saw it N days ago. Two scenarios:

1. You don't delay. You review the card now. It's either hard and reset to 1 (see it again tomorrow) or easy and you see it again in 2N days.

2. You delay. You review it M days from now. When you finally do it's either hard and reset to 1 (see it again the next day) or easy and you see it again in 2(N + M) days.

That's it. The algorithm has you covered if you delay. It doesn't do something silly like say "This card was supposed to be reviewed after 2 days, but you waited a month. You remembered it, but we're going to show it to you again in 4 days." The algorithms will take the delay into account (maybe not one-for-one) like I illustrated above.