In the absence of any official article, here are my current house rules:
Include phenomenon cards in the planar deck, as normal. As a rule of
thumb, the deck should include at least 25 actual planes.
Players "encounter" phenomenon cards when they are revealed from the planar deck -- put their triggers on top of the stack immediately.
In most cases, this means that the phenomenon will be adjacent to
the current plane. After resolving a phenomenon, place it on the
bottom of the planar deck and continue revealing plane cards until all
spaces are filled as normal.
Spatial Merging gets errata to read "Place both of them in the
same slot on the planar map" instead of "Simultaneous planeswalk to
both of them."
My main guiding principle for these house rules was that phenomena are random, one-time events, so it was aesthetically unappealing to have them sitting in a spot, waiting to be triggered. Because they activate when revealed from the deck, moving to a known square can still have additional unforeseen consequences (e.g. Morphic Tide or Chaotic AEther). I like that phenomena only occur when exploring the edges of the map: they are surprises that afflict planeswalkers venturing out into unknown reaches.
Overriding the unique "encounter" keyword allows all current phenomenon cards work as written, with the exception of Spatial Merging (and their reminder text). I consider this an elegant solution.
Phenomena are placed at the bottom of the deck (rather than shuffled) because I found that more natural and expedient. They can still recur unexpectedly because the deck will be shuffled whenever out-of-reach planes are removed from the map. You could also shuffle them in immediately if you wanted to; which approach you prefer doesn't really matter.
Note that, with too few actual plane cards in the deck, the game could get into a degenerate loop when you continuously reveal phenomenon cards -- hence the warning about minimum deck size. Right now, we play with a deck of around 35-40 cards total, 5 of which are phenomena, and this has worked well.
There are some niggles here with trigger ordering that I haven't investigated fully, but so far my informal play procedures have been sufficient for a few quick playtest games.