Some lasers will indeed get less efficient if they're too cold.
In some systems, you may want to keep the laser at a fixed operating wavelength, and this requires operating the laser at a fixed temperature.
Cooling the laser below 0 C could cause ice to form on the laser (if it's not in a hermetic package), impacting performance and reliability.
If none of those apply, excessive cooling will simply consume power unnecessarily, increasing the power consumption and the demand on the cooling system for the whole system or plant.
A TEC controller can be as simple as a 6-pin microcontroller, so it doesn't add much cost to the system, after you've already budgeted for the TEC module itself and the components to drive it efficiently, so even if the benefits are small it might still make sense to add the controller.
The controllers I am talking about are ones like the Newport 325 which can cost as much as $1,200 new.
The $1200 mostly pays for the display and user interface and a remote control interface (GPIB, USB, or ethernet), as well as the amortized cost of R&D to develop the design and its software, NRTL listing, etc.
You'd buy one of these instead of building your own if you're only building a small number of systems (maybe only one). $1200 is less than the cost of designing, building, testing, and debugging a new controller design, even if the cost of materials for the new design approaches $0.
Many laser users (say, biology researchers, or even electrical engineering researchers in other specialty areas) may not even be skilled in PCB design, control systems, embedded programming, mechanical engineering, and other skills needed to develop a controller on their own.