Looks like barcode information associated with that read. Sounds like you have the same type of data as reported in this topic and it sounds like that user did not need to trim prior to downstream analysis.
Having dual-index barcodes in the header line is not really a common format so we do not have a way to trim or explicitly process this information in QIIME 2 yet. Fortunately, your data are already demultiplexed so it sounds like this should not be a problem. If it is, it would be simple to trim, e.g., with a bash script.
Whether they overlap depends on the length of the total amplicon and the length of your reads. You should have information on both of these. You need a minimum of 20 nt for successful read joining.
Check out this tutorial and this tutorial for more details. You will use the output of qiime demux summarize to assess read quality profiles, which will inform trimming strategies.
Your sample metadata file contains all information about your samples — e.g., sample ID, time collected, treatment group, patient information, etc. Take a look at this file from the tutorials to get an idea of what a metadata file looks like and what information it can contain. See this tutorial for more information on file formatting.
I hope that helps!