Is the block 'Inner Loop/e^(-s)' driving the block 'Inner Loop/Sum'? It looks like the 'e^(-s)' block is trying to set the Sum block to be double, but the Sum block is already set to some other data type. I'm not sure why that's happening, but here's a snippet from the help for the Sum block documentation,
Inherit: Inherit via internal rule
Simulink chooses a combination of output scaling and data type that requires the smallest amount of memory consistent with accommodating the calculated output range and maintaining the output precision of the block and with the word size of the targeted hardware implementation specified for the model. If the Device type parameter on the Hardware Implementation configuration parameters pane is set to ASIC/FPGA, Simulink software chooses the output data type without regard to hardware constraints. Otherwise, Simulink software chooses the smallest available hardware data type capable of meeting the range and precision constraints. For example, if the block multiplies an input of type int8 by a gain of int16 and ASIC/FPGA is specified as the targeted hardware type, the output data type is sfix24. If Unspecified (assume 32-bit Generic), i.e., a generic 32-bit microprocessor, is specified as the target hardware, the output data type is int32. If none of the word lengths provided by the target microprocessor can accommodate the output range, Simulink software displays an error message in the Simulation Diagnostics Viewer.
You can try forcing the output data type to be double, if that's what you really want, or you can try putting a Data Type Conversion block in front of the Sum block. One other thing that can help is to try turning on Port Data Types from the Format menu. It should show you all the propagated data types when the error happens.