Probably the most important aspect of the ZL/ZR issue is to recognize that these conditions lie outside of the engineering standard envelopes used for icing certification. These conditions fall into the category of SLD (Supercooled Large Droplets) .They are conditions that were not considered during the design, testing and certification of the aircraft ice protection systems, and the aircraft handling characteristics were not evaluated with ice shapes derived from ZL or ZR.
The engineering standard envelopes use a single-mode, bell-shaped droplet size distribution which is centered on a median droplet size of 40 to 50 microns. Under the far right hand tail of the curve, some ZL-sized droplets may be found, but they are extremely sparse. The vast bulk of the droplets in this standard envelope fall well under the size of ZL. This engineering standard is believed to cover 99% of the icing environment, although that number was originally based on the entire scope of icing in the continental US. It does not accurately describe such places as the Great Lakes region, the Canadian maritimes, or the Alps of northern Italy.
Like most design choices, the extent of the protected area on the wing is an engineering optimization. One common method for determining the aft extent of the protected surface is to determine how far back a 50 micron droplet will impinge. There is no requirement for protection further aft because the larger droplets that will impinge there are not part of the envelope and are considered quite rare. Some manufacturers will opt for more conservative choices; the 50 micron drop estimation is part of what got ATR and Embraer in trouble with their turboprops. ATR resolved this in the weeks following Roselawn with a retrofitted set of wing leading edges which extended the protected surface further aft by around 1% MAC.
The industry is in the process of developing certification standards for a new icing envelope which is centered on large droplets. No airplane has yet been certificated to this new standard; it remains to be seen just how it will work.
Although ZL and ZR are thought to be rare, they actually make up 1.8% of the reported surface precipitation in the continental US. In the icing accident database that I maintain, ZL or ZR are associated with approximately 30% of the accidents and incidents combined. This makes a pretty strong case that SLD is a very serious threat to safe operations. As with all icing, aircraft scale plays a significant role, with larger transports not experiencing the same degree of difficulty. However, the 30% number remains pretty consistent up through the large turboprop scale.
In terms of vertical extent, an extensive study of atmospheric measurements done by Dick Jeck at the FAA (now retired) identifies ZR environments as deep as 7000 feet and ZL environments as deep as 12,000 feet. I have a couple of reports of crews making altitude changes in excess of 4000 feet and not escaping the condition. That said, most of these environments are shallower and an altitude change is definitely in order when they are encountered.
While ZR generally exists under an inversion, ZL can exist without an inversion through collision coalescence. One fairly good method for anticipating ZL or ZR aloft is the presence of ice pellets at the surface.
The elephant in the living room for the past sixteen years has been the existence of holdover tables for light ZR and light to moderate ZL. As far as the FAA's Aircraft Certification service is concerned, the aircraft is not certificated for these conditions and should not be operated in them. As far as the FAA's Flight Standards Service is concerned, they are an acceptable norm of all-weather flying. This makes it almost impossible to generate a consistent, factually based message. When taking off in moderate ZL, for example, the fluid work quite well within the holdover times. Once the airplane rotates, the fluids shear off, and one is suddenly airborne in an environment that the airplane's ice protection system was not designed to cope with. For years, this was tolerated on the premise that the SLD layer was very shallow; this is not at all true in a reasonable number of cases.
There are still no real good, black-and-white answers, which is what makes the icing issue such a squirrel. Scale is important, but you can expect some performance degradations even in larger scale aircraft. Proactive operation of the ice protection system is essential.
Here are a couple of links that may provide some good references:
http://www.tc.faa.gov/its/worldpac/techrpt/ar0945.pdf (Dick Jeck's paper)
http://www.skybrary.aero/bookshelf/books/531.pdf (AIAA paper on the icing accident database)
http://www.skybrary.aero/bookshelf/books/532.pdf (AIAA paper on the history and nature of the airframe icing threat)
http://flightsafety.org/fsd/fsd_jan96.pdf (John Dow's excellent article on roll upset)