It applies to anything that can send any SIP packet to Asterisk (as well as anything to which Asterisk can send a SIP packet)
I think chan_pjsip defaults to the first configured transport, in that case.
SIP has no concept of trunks, so PJSIP has no concept of trunks. As such a local device and a “trunk” are no different at a basic SIP level, although, typically trunks support multiple calls (that’s where the term comes from) and expect to have dialled digits in user part of the URI
Also, a common recommendation is to not use port 5060, to reduce the number of toll fraud attempts that you receive, so knowing how to change the bind port is important.
Yeah. There is your problem. They both match to the same source IP. That always wins. So you need to stop that. I still dont know why you need two trunks. There are better ways to deal with billing it.