Let’s make a few guesses at how it works:
The banks have some bad assets on their books and are short of capital. They pay an annual premium to the government to cover the banks on any losses above a certain amount on those bad assets. Let us say the figure is 50%. So the banks take the first 50% of losses/risk and the government take the risk on the remainder, which is hopefully an acceptable deal for the tax payer - but don’t count on it.
The advantage for the banks is that with the government guarantee they may expect to be paid out to a limited extent if the assets go bad (although don’t count on it because of the state of the UK economy), but to the extent they are covered by government guarantees they will be zero-rated for capital adequacy purposes. So the banks will pay a premium of perhaps 1% of the amount guaranteed and will thereby avoid using up Tier 1 capital equal to 8 or 9% of the risk adjusted amount that they would otherwise have to pay if/when the bad assets are downgraded. Not a bad deal for a bank that is short of capital.
On the other hand the government gets paid a premium that it pretends represents value for money and in return incurs a contingent liability. The government likes that because not only does it have to come up with any more cash for funding (which needs more borrowing), but it also gets to exclude the liability from the government accounts (on the same basis that it excludes the guarantees it has given against National Rail's borrowings). There is no valid accounting reason to do so. It is simply using its often practised techniques of lying and self-delusion.