It should be obvious that an attempt to use any numerical rule such as the 10/10 test will quickly run into problems. Suppose a contract has a 1% chance of a 10,000% loss? It should be reasonably self-evident that such a contract is insurance, but it fails one half of the 10/10 test. It does not appear that any "brightline" test of reasonableness nor signifance can be constructed.

Excess of loss contracts, like those commonly used for umbrella and general liability insurance, or to insure against property losses, will typically have a low ratio of premium paid to maximum loss recoverable. This ratio (expressed as a percentage), commonly called the "rate on line" for historical reasons related to underwriting practices at Lloyds of London, will typically be low for contracts that contain reasonably self-evident risk transfer. As the ratio increases to approximate the present value of the limit of coverage, self-evidence decreases and disappears.

Contracts with low rates on line may survive modest features that limit the amount of risk transferred. As rates on line increase, such risk limiting features become increasingly important.

Reference: www.wikipedia.org