The results of your simulations for various approaches to insurance cover are very enlightening. I like the "light cover" approach. Here's another candidate.

Simply take insurance whenever the amount of your bet is equal to or larger than the amount that corresponds to the true count at which insurance should be taken.

For example, with the Zen count and a 6 deck game, insurance should be taken at true counts of 5 and above. If your "target" bet for a true count of 5 is, say, 3 units, you would take insurance whenever your bet was 3 units or more.

After all, this is how insurance, in the traditional sense, is supposed to be used. Retain the more frequent, less costly risks, and buy insurance to cover the less frequent, but more costly risks.

Of course, the constraints imposed by normal cover betting will cause some losses when this approach is used. Sometimes the count will reach the level at which we should take insurance, but, due to parlaying lag, our bet may not have reached the target amount, so we would not take insurance. Similarly, the count can deteriorate, but we may not have managed to scale our bet back yet. So, to maintain cover with respect to our insurance betting, we would take insurance even though it is a negative expectation bet. Can you determine how these losses would compare with the losses you have quantified for the approaches you tested?