I wouldn't recommend using sensors with different laser frequencies either, because that likely means using an entirely different product, with different accuracy and other properties.
don't tilt, shift
I think the idea of tilting a laser scanner is very problematic, especially with a height constraint like you have. But what about shifting the scanner up or down in each unit?
Let's say if a scanner does not interfere with another one if it's 1 cm above the other. You can give the scanner on each robot a unique height they can operate on without causing any problems.
This solution is limited to a small number of robots, depending on the height difference necessary between two sensors and the available height on your robots.
Given that you have established communication between the bases, each one could do its scan while the others are not using their sensor. After the scan is finished, a message is sent to the next robot.
This reduces the scanning frequency as they cannot happen in parallel. Depending on the range of the laser scanner, the size of the area and distribution of robots, this could be compensated by the fact that each robot scans the other one as an "obstacle". Each robot could estimate the new positions of all robots it sees and message that information to others after it did its scan. This can be tricky when it comes to identifying the other robots from the scan and if one robot could occlude another one.
shared pose knowledge
If each robot knows the poses of all other robots, it can exclude sensor readings from those directions. Depending on the situation, careful sanity checks might be necessary before excluding sensor data.
Otherwise an obstacle that's exactly between two robots would be "seen" by either one because sensor values detecting it were discarded for being "in the shadow" of the other robot.
give it a try
I'm not exactly sure how those scanners work in detail. Does both the laser and sensor rotate? Or are they omnidirectional? Or is either one omnidirectional?
Maybe having all sensors on the same height isn't that much of a problem unless two of them face each other exactly, which appears to be unlikely and even in case of failure would only falsify a few measurements. With the strategy of shared knowledge of the poses of the robots mentioned above, it should be possible to filter those measurements out.
This is not some DIY hack, but product sold by a competent company. If in doubt about possible problems with their products, contact SICK and explain your project in detail with all the requirements and the issue that could arise. I'm sure they are willing to help you with their expertise given that you want to buy (or have bought) several sensor units from them. Even if not, it's definitely worth a try.