Damage risk if 5v comes in on analog read pin when set to INTERNAL voltage ref?

I need to be able to measure resistances in the 0.1-10ohm range. My configuration is working well, but it leaves a little room for improvement in resolution.

+5V ------ 40kohm ---+-- 10kohm ------ 0V
                     |
              test_resistance (can float from 0ohm to infinity)
                     |
                     0V

If I set the analogReference to INTERNAL and measure at '+', a value of 3 or 4 indicates a test resistance of about 1.5ohms. This is good sensitivity, but if I could tolerate a much higher voltage (say, 4.5V) on that input pin while using the internal analogReference, without damaging the arduino, I would replace the 40kohm with a 1kohm and see a 4.5x increase in resolution.

Safe or Stupid? (Or... "Who let this guy on the Internet?")

Sorry that diagram doesn't make sense. Where are you measuring things? As drawn you have the test resistor in parallel with a 10K forming a potential divider with 40K. By the way where do you get 40K resistors from, it is not a standard value.

Yes, my test resistance is in parallel with the 10kohm. I'm measuring at the '+'. The 40kohm resistors aren't standard - they come in the same packaging as 10kohm resistors and require some assembly! (It's what I had around, and it seemed a simple - if cheezy - way to test out my ideas before buying the resistors I would need to do this with only two resistors at a 4-to-1 ratio instead of 5 10kohm resistors.)

Is my test resistance sinking current fast enough that its destroying the quality of my measurement? I figured the voltage divider would give me 1V to work with to compare with the internal voltage reference.

I don't really need to be able to distinguish between .1ohm and .2ohm, but it would be bad if I didn't have at least 10 levels of input between .1ohm and 2ohm.

I'm measuring at the '+'.

So the + point is the analogue in. That makes more sense. However the 10K resistor does not make any sense. Why have it? You can put any voltage into an analogue input up to the supply voltage of the chip. It is fine to put in a voltage higher than the reference voltage, you will just get a reading of 1023. Just so long as it is 5V or under.

Is my test resistance sinking current fast enough that its destroying the quality of my measurement?

I have never come across the concept of fast current. What do you mean?

I just realized where I blew it. I'd seen a similar setup that required much lower resistance, but by cranking up the resistance to bring down the current and keep a more stable voltage reference, I lost any real accuracy in my measuring, comparing the resistance of my test load against that last leg of the voltage divider. I guess I should just reduce the question to "If I feed more than 1.1V, but less than 5V to an analog input pin that is internally referenced, can I do it any damage?

"If I feed more than 1.1V, but less than 5V to an analog input pin that is internally referenced, can I do it any damage?

The answer is no, you do not do any damage.