In Ark, all virtual tx outputs (vtxos) are actually outputs on the leaf txs of a binary tree of transactions. The reason for this is two-fold: so that a single person exiting doesn't have to broadcast all the vtxo outputs for all vtxos in the Ark tx and so that in the case of mass exit, the process smoothens out and there is more time for everyone to exit slowly.
You could think of the vtxo tree of a tree of transactions that each have one input and two outputs and like that they form a tree structure. Think of these transactions as transactions that are already pre-specified upfront. They can either be pre-signed or pre-specified using introspection opcodes. This means that all these transactions are valid Bitcoin transactions and could all in theory be confirmed. However, they don't carry any fee and no one has an incentive to broadcast them to the network.
Such a tree might look like this:

In this tree, there are 4 leaves, A, B, C and D; they are actual valid transaction outputs and we will call them "vtxos". Virtual because they are not confirmed yet, but can be confirmed if they would be mined.
Now, for the vtxo A to be confirmed, let's say because its owner wants to spend the money on-chain, all the transactions going from the on-chain output (the top-most "cov") towards the leaf A have to be confirmed on the chain.

Because this is a binary tree, the number of nodes in a branch is log2(n)
where n
is the number of leaves, plus one for the leaf itself. So that means log2(4) + 1
is 3 transactions to be confirmed.