Opcodes beginning with Hex 'A' (1010_xxxx_xxxx_xxxx) and Hex 'F' were traditionally used on the 68000 architecture to implement coprocessors.
For example, the 680x0 family originally had a separate Floating Point processor. All opcodes for floating point operations had an 'F' as the top nibble, and the 680x0 handed them off to the coprocessor for execution. If you tried to execute floating point code on a 680x0 processor without an FPU, you would get an "Unimplementd F-Line Opcode" exception.
The ColdFire processors use 'A-line' opcodes for MAC instructions, and still use 'F-line' opcodes for the FPU (where present). If a particular A-line opcode doesn't decode to a valid instruction, you get an A-Line exception.
On the old 68K processors, A line opcodes (so called because the first bits of the instruction were 1010, or hexadecimal 'A'), were instructions that would cause a jump to a special handler. The handler would look at the remaining bits in the A-line opcode and use those to invoke different handler routines. Most often the A-line opcodes were used to implement a program's calls to an operating system. The 68K-based Mac OS used A-line opcodes this way. I'm not familiar enough with the Coldfire to know if it has the same feature.