SWR for Solid State Linear - Question

This next question is more abstract.  I hope you can help me unravel the puzzle.

Before I tackle the big QRO solid state linear, I'm building a smaller solid state linear.

Let's assume one of these MOSFETs:   IRFZ24PBF    In later designs for QRO, I'm going to get one of the monster NXP power MOSFETs.  But, before the QRO, I want to establish some framework to test, hence the point of the question below.

Now for the setup:

Let's think abstractly. 

Suppose I have two "devices" (home brew etc.)

One is a "SWR detector".  It is a device with two inputs and one output.

The two inputs are RF_IN_SENS, and RF_OUT_SENS

The one output is N_TX_ENABLE

When the "SWR detector" (analog circuit) examines the RF_IN_SENS and RF_OUT_SENS it will pull N_TX_ENABLE (active low) when the SWR exceeds a redline (a redline designed in the SWR detector).

When RF_IN_SENS compared to RF_OUT_SENS reveal a SWR within limits, N_TX_ENABLE is pulled high.  It's an active low signal.

Push that aside.

Next the the other device is "RF Amplifier"  and reduced to simple terminals has:

TWO Inputs:   KEY_DOWN  and RF_IN

THREE outputs:  RF_IN_SENS, and RF_OUT_SENS, RF_OUT.

Assume that the amplifier samples the input and output for the benefit of some external equipment, The SWR Detector.    Hence RF_IN_SENS, RF_OUT_SENS -- those are samples of the actual signals.

The raw signal from the exciter is given to RF_IN, and the raw amplified RF is delivered by RF_OUT.

Here's the question:

Suppose the SWR detector pulls N_TX_ENABLE low (it finds bad SWR), active low.

I do not know if I should:

1. Trigger that the RF_OUT is attenuated to a low power -- this keeps the "bad SWR" present, but lowers the RF_OUT so that while the SWR is bad, adjustments can be made (safe for the finals) to alleviate the bad SWR.
2. OR, trigger that the RF_OUT is shunted off to a perfect load to protect the finals while the SWR can be fixed.

The problem I have is if I choose #2, I lose the actual bad SWR (even if RF_OUT is attenuated) to know when the Detector has been satisfied.

If I choose #1, I lose the opportunity to 100% protect the finals from any bad SWR, but I lose also the actual bad SWR condition to know if I've corrected the problem.

Eg, in other words, if the SWR detection exceeds the redline my instinct is to let the Amplifier see the bad SWR, but just attenuate it sufficiently to protect the finals while the tune on the antenna load can be adjusted. 

Attenuate or cut off completely?

I assume also that it's right to attenuate the RF_OUT, not RF_IN while the redline condition is detected.

I'm asking this because when I build the small-scale (60W SS linear) I want to build up a test-framework that I will want to use when I pivot to build a bigger SS Linear.

If this doesn't make sense OR you're too busy to entertain, that's fine.   Let me know and I'll ponder it more.   Worst case scenario is I melt some transistors.

Comments this question are welcome.

Remember, it is in the abstract.