The question is not easy to answer as it would be difficult to give exact print guidelines. This answer from user @typo already shows results of the print height versus the specimen strength (IMHO his answer should be the accepted answer), as taken from this excellent reference at 3DMatter which basically describes the results of a series of experiments. This answer builds upon his answer. In reference to your question, this reference did not optimize the print settings (all specimens are made with the same print settings), so your question is valid.
It is assumed that you imply in your question that all print parameters that effect the inter-layer bond strength needs to be taken into account for the optimization. Many parameters are in play to bond the filament onto the previous layer, amongst these parameters are e.g.:
Print speed
Filament temperature
Print height
Print cooling
All these parameters influence the deposition temperature which in its turn determines the bonding to the previous layer. Not only printer parameters play a role here, but also the properties of the material itself. Between the various brands, and even within a single brand, material variations (e.g. color doping, or different process batches) influence these parameters.
The question states
It seems obvious that too thick a layer will give less compression and maybe less heat transfer into the layer below
Well, this is not so obvious and assumes that compression is the main driver for bonding a layer. However, the larger the layer height, the more filament can be deposited at once with a higher heat capacity (stays hot longer), so potentially this could have a positive influence on the bond (higher temperature, better adhesion).
The print fan cooling parameter could play a very important role here (or even the filament print temperature). In fact, the results of this are already shown by the 3DMatter experiment referenced above. If you keep all print parameters the same except for layer height, the bond strength increases. This implies that in order to get a better bond for low print heights, you should decrease the amount of print-fan cooling flow, or increase the filament print temperature. How much this is should be done in a similar experiment where you lower the cooling air and increase filament print temperature for more specimens (separately) and test again. This is referred to as a design of experiments.
Theoretically, you can make the bond at any layer height just as good provided you optimize the correct parameters. This implies that there is no relation between the print height and the bond strength, it is just a matter of proper setup.
Also, I would not say that poor bonding strength is the cause of failure in Z direction, as FDM deposits layers in between each layer you will have a lot of potential crack initiation locations, this is usually the starting point of the failure. I have seen prints start the failure between the layers, but not continue to crack along the layers, but traversing through the layers meaning that the bonding strength is not that bad after all.
