The things that break are the things living in one head. A supplier who only deals with you. A process nobody else can complete. An account only you can access. A customer who expects you personally. None of these are visible during normal operation, because you resolve them continuously without noticing you are doing it.
Do it properly rather than symbolically. Genuinely unreachable, for long enough that the ordinary rhythm of the business has to continue without you, which usually means at least a week. Checking messages twice a day means you are still operating and the test tells you nothing.
Prepare only what is necessary: tell customers, set an honest out of office with a return date, and make sure somebody can reach whatever is genuinely urgent. Resist the urge to prepare thoroughly, because thorough preparation is how you accidentally do the work in advance and learn nothing.
Then write down what broke, because that list is your year two agenda. Each item is either something to document, something to automate, or something to hand over. Working through it is considerably more productive than any planning exercise, because it was produced by reality rather than by speculation.
Businesses that never run this test still find out. They find out during an illness, a family obligation, or an emergency, at a moment they did not choose and with no capacity to respond. Running it deliberately in a quiet period is the same information at a fraction of the cost, and the recovery is a benefit rather than an expense.
If a full week is genuinely impossible, run a shorter version rather than skipping it. Three consecutive days where you are unreachable will surface most of the same dependencies, and it is considerably better than the alternative of never testing at all. What matters is that the interruption is real rather than that it lasts a particular length.
Do it in a quiet period rather than waiting for a convenient one, because a convenient period does not arrive. Businesses that defer this until things calm down defer it indefinitely, and the test is most useful precisely when there is enough activity for something to actually break.
The secondary benefit is worth naming. Returning from a genuine break produces better decisions about the year ahead than any amount of planning conducted while inside the daily work, because distance is what makes structural problems visible. The planning you do afterward is usually better than the planning you would have done instead.