Professional sensors can use TDOA or combinations of TDOA and other techniques to plot the source of a transmission at significant distances to very small areas. I have used TDOA and other passive systems capable of plotting the position of a target at 10's of km to within a few meters, at hundreds of km to within a large city lot, and at thousands of km to within a few miles.
I worked on some tech that, once I was allowed to see how it worked and the results, was very surprising to me. That was the type of stuff that if I were a pirate, I might worry about. That was decades ago. Kiwi TDoA is not as good as what I was seeing then and consider that the world has moved on since then. You get the idea.
The multi transmitter solution has the most probability of success, as long as they are all the same mode. What is that going to do to the audio though? Propagation delays will produce echoes, even if the audio starts out in sync.
Since we are talking about potential countermeasures, here's some spew of (un)conciousness from me:
1) Perhaps one way would be to set up a triangle or polygon of low-power "confusion transmitters" in the ground wave area around the real TX, so that when they actually come near you to triangulate your exact location, there is some error introduced into the determination. Throw in some random amplitude and phase modulation onto the confusion transmitter signal. The effectiveness would greatly depend upon many variables. Think of this as "local jamming" or akin to a fighter aeroplane throwing out heat-producing chaff to confuse the infra-red detection systems of oncoming missiles. Obviously this would be wholly inappropriate for low-powered, non-HF pirates where local reception is desired.
2) Here's a real brain fart: if you could have one TX producing NVIS (
Near
Vertical
Incidence
Skywave, bouncing off the F2 layer and coming straight down) and simultaneously another TX producing a "normal" (non-vertical) skywave signal, this might introduce enough "confusion" local to the TX. However: a) finding a single frequency where both NVIS and normal skywave work simultaneously would seem fundamentally impossible. b) It might be unpleasant for many listeners. (See Token's comments above.) Like I said, this is a brain fart.