So, to learn something, or just for S&G, I did a bit of reading about self-heating of an I2C device because of activity on the bus and couldn't really find anything at all (ok, I did not look that long). There is a decent discussion on the matter of over-heating in a TMP116/117 sensor.
https://www.ti.com.cn/cn/lit/pdf/snoa986 But, the discussion concerns the size of the pull up resistors and their location near the chip (drawing current and producing heat), and frequently accessing the chip (increasing the current draw/per unit of time) - nothing about activity on the bus for other chips. Now, if you had two sensors right next to each other and frequently activated one of them, you could heat up the other externally...at least in principle.
But heating up the one because of any activity on the bus seems pretty far-fetched to me, especially for something like that to be mentioned in *that* data sheet, but I could be wrong. Don't know and I still tend to think that if there was a mis-translation it probably had to do with warning about two (or more) devices using the same address.
In any even, while I am unconvinced about self-heating to the extent of decreasing because of any other activity on the bus. Nevertheless, I am glad that you contributed your observation of using two devices and changing the address of one and having no apparent problems.