The scenario you have described is all OK.
Nevertheless, we recommend that you enable GTID on all nodes, and the recommendation in that reference you provided is also to enable GTID on all nodes. As a general principle, Percona recommends having the same configuration on all nodes of the cluster. If you don't do it at the same time and then later try to enable it then you risk having GTID inconsistent as per that reference you posted.
As long as that recommendation is understood...
If the node that is being shutdown is replicating trx .. then existing trx will not have any affect
If the node that is being shutdown is processing an active workload then you should stop this workload. If the trx is already in commit state that will processed but no new trx will allow to progress.
There is a discussion on the Percona blog that describes some different scenarios using 5.6 and this might provide some further insight for you. For example, it observes that even if you are using 'mixed' replication "we can have different replication modes on different servers, but only if all servers have gtid_mode set to ON."
However, the post also mentions that the author can't think of a good reason to do this (maybe you now have one). As well as blog posts (do a search on that site) you might also find some useful - free - recorded webinars on PXC that might be good resources for you and provide more information. Here is the link, as always please read the details carefully and test! :)
Here is a webinar for 5.6 https://www.percona.com/resources/mysql-webinars/whats-new-percona-xtradb-cluster-56 ... if you add webinar to your site search you'll turn up some more useful recorded webinars on Percona XtraDB Cluster. These are for techs and presented by techs, normally support or development engineers.
Disclosure: I work for Percona (obviously, but I guess I should say it anyhow).