I think you might be considering a different situation @ccsh. I believe you might be thinking of the case where both RX channels are receiving the same signal on two antennas through an open air channel. In this case, yes, your phase offset between the 2 RX will vary randomly. But for both @will_mysky and @Aridane, the situations they are describing are much more predictable conditions - described here and here, respectively. In the case of a direct connection through a splitter or a coupler, like described in those links, the phase offset between the two channels should not vary, seeing as they are running from the same PLL. In essence, it is very similar to how you would measure the properties of a device using a VNA. But despite this, something is not as it should be or something is just not clear yet, seeing as the LimeSDR currently doesn’t work in that way.
I had this same issue some time ago, but haven’t had time to get back to trying to resolve it, until now. It would be really good to get some feedback from the LimeSDR design team on it, as it doesn’t seem to simply be a matter of turning TDD mode on or switching off the DC corrector. The problem I had was able to be circumvented by using the loopback as my second channel, seeing as I just needed a phase measurement - but that option is much, much slower than using the two channels together, as switching the loopback on and off is not a speedy operation.