From my personal research experience...
Acoustics studies of marine mammals are complicated by the fact that humans only visually observe a small fraction of their behavior while the animals are at the surface and it can be difficult to acoustically discriminate vocalizations between species in mixed-species encounters.
Many years ago I wanted to know if mixed species schools were more/less likely be to detected using a towed hydrophone during closing-mode shipboard visual line-transect surveys in the Pacific Ocean. These were population-level surveys and my question was related to detecting the presence of groups of dolphins based on detecting any combination of whistles/burst pulses/echolocation clicks (I needed to know more about groups that were NOT vocal!). I was unable to reasonably determine if which species was vocalizing, but instead I looked at the overall acoustic detection of each group. As these sightings were associated with visual observations and group size estimates, I used group size as a co-variate. The overall results were that mixed species schools were consistently more vocal than non-mixed species schools, but that mixed species schools also tended to be larger group sizes. Also, quiet dolphin schools tend to have smaller group sizes in general. There was a bit of variation by species and in some geographic regions. These results were presented at an Acoustical Society of America meeting in 2005 (https://doi.org/10.1121/1.4787441), but were not published.
These are fairly simplistic results-- but using a lot of good data from visually confirmed sightings across an incredibly large geographic area. In the intervening 17 years (!!) our ability to classify dolphin acoustic detections has improved dramatically (it was essentially non-existent at that time!). Eventually as our ability to correctly classify acoustic detection approaches 100%, we should be able to (hopefully!) identify the presence of multiple species in mixed-species groups. Classification methods such as BANTER (see publication here and package here) that allow for assessing the classification scores for acoustic detections may provide a path forward for detecting/classifying mixed species groups (see PlotVotes under 2.4.2 in this BANTER tutorial).
Once we can identify the multiple species in an acoustic encounter--we then have an improved capacity to apply this to scientific questions. Of course, this example I've shown here is for an encounter-- looking at specific calls is even more complicated.