Cool!
Something doesn't add up here. The V3-V4 region is too long to be merged with a 2x150 Illumina run. Are you sure about these descriptions? Your stats results below indicate that your reads are merging just fine, not possible with 150 reads of V3-V4.
I don't know much about invertebrate gut but I seem to recall that diversity there is lower than mouse/human gut. So the 3167 features you described might not necessarily be low at all, unless you have evidence from previous work that more features are expected. Remember these are the # of unique features you discovered across all your samples, not to be confused with total # of reads that the dada2-stats shows.
Your dada2 stats results actually look pretty good to me and are on par with what we'd expect. You have loads of reads 50k+ to continue with your analysis without worrying but if you really wanted to try and improve your results we would need a bit more information. For example uploading the demux.qzv as well as your dada2-denoise stats vizualization artifact (not just the image) here. The artifacts hold important information in their provenance tab that can help us explore your approach so far.
In the future please try to avoid asking multiple unrelated question in the same thread. This helps us keep the forum organized and easier to search the archives for past questions (including this one has been answered before).
I would recommend asking your sequencing facility or whoever did the amplicon preparation for this information. It will be required for publication anyways so you might as well get it. If for some strange reason this information is lost, you can always use a generic classifier that isn't trained on your specific primers, albeit with slightly less accuracy. This is available on the data resource page.