In my house the AC compressor is located at the place where I was planning to install the inverter. The AC uses 240V and has 2x30A circuit breaker on the main panel. The AC unit is connected to the "wall" by a simple disconnect panel, which as room to install second conduit for inverter.

The benefit of this is that I do not need to route any new wire for the inverter or install new breakers for the main panel.

On the downside there may be some surges/sags on the power line when compressor starts/stops. Don't think these are too bad to cause damage to the inverter but could cause it to disconnect from the grid (and wait that 5 minutes).

Just a question that is this a crazy idea to start with?

I'd say yes:

1) It sounds like you are thinking of wireing into the compressor circuit?
that may not be good, as you need 2 breakers for the inverter, you should not share the AC's compressor breakers.

2) you really don't want the inverter to cycle each time the compressor does.

3) mount it where handy, and run conduit to the main breaker panel. That leads to a no-questions inspection by the city

The inverter might trip offline when the AC kicks in, as the voltage will drop at the end of the line at the compressor for a few cycles for sure. So every time your AC klicks in, the inverter might drop offline for 5 minutes as UL1741 requires when an AC paramter is out of spec.

As mike points out, you have breaker/code issues as well. The Inverter needs to be on it own, breaker and line back to the main panel.

To better understand why this is an issue and not just some code thing, One example is the AC is no longer protected by the 30 amp breaker, as the inverter can supply say an additional 15 amps, so from the AC's view, thats now 45 amps which is clearly over its required over current protection.

Now you could replace the AC disconnect with a similar sized sub-panel , say a 6 circuit one and be NEC ok, but then you have the drop out issue, the inverter, will actually help the AC load and minimze the voltage drop, you would just have to wire it up and see if its an issue.

That arrangement may very well cause line sags severe enough to interfere with the inverter. But if it worked, I might try what you just suggested installing additional breaker panel.

Just out of curiosity how fast the line voltage detection of typical inverter (like Xantrex GT3.0) is? It must have some filtering to reduce false tripping due to sudden spikes... but does it respond with 8ms (half cycle) or 16ms (full cycle), or even longer? How is the RMS measuring implemented? If it is just a simple averaging with long time constant RC filtering the delay must be very long... Is it sensitive to harmonic distortion, such as caused by low Power Factor appliances connected to the grid?

I'll bet starting a compressor motor will sag the line for a good 5 full cycles, if not more. Kind of depends on the drop from the power company, and the wires to your panel.
You may want to feed your inverter wires into the compressors J-box, and if the condiuit is large enough, thread the inverters wires back to the main panel, saves running a new conduit, if the existing is large enough to meet code. I'd use very good rated wire, if the compressor runs very long, it's wires will warm some too.

The inverter WILL kick out, a loaded motor start will pull the line down many cycles, typically 10 to 30 cycles, and also cause a large power factor shift as the motor accelerates, starting is mostly an inductive load.

Taking the inverter back to the main panel is the best route to go.