(...)
If by event horizon we understand the proper distance* from beyond which no signal can ever reach us, even after infinite time, then that distance is not shrinking - it is and always will be growing, albeit at an ever slower pace, asymptotically approaching 17.3 billion light years. At least that's what the best-fit models tell us.
Here's the evolution of the event horizon in the currently favoured model plotted on a graph (the outer thick solid line is marked 'event horizon'):
View attachment 94790
However, galaxies (the dotted lines on the graph) are constantly carried away beyond this horizon by the expansion. This means that every second there is less and less matter that we will be able to observe in the far future as it is now. I.e., today a galaxy on the edge of the horizon might emit light that will eventually be seen far in the future, but the light that it emits tomorrow might never reach us, so it will never be observable in its state as it will be from tomorrow onward.
In the far future, you'd end up with a universe where all but the closest of galaxies will have left the observable universe. So (...) galaxies vanish from sight one by one.
Only they don't really, at least not in a straightforward way - due to the fact that the closer an emitter is to the horizon, the more stretched its light becomes, and the longer it takes for it to reach us (with light emitted
at the horizon requiring infinite travel time, but still getting there), it will be possible to observe, if not now, then some time in the future, any galaxy that was inside the event horizon at least once (but the farther away a galaxy is, the less 'up to date' the picture will be). This means that everything up to the current(!) proper distance of 63-ish billion light years will be eventually observable, and everything already observable will always remain so, but in increasingly outdated and redshifted state.
Only not even that is entirely correct, as the increasingly more and more stretched wavelengths of light emitted by objects near the event horizon will eventually get stretched so much that not even a receiver the size of the observable universe would be able to detect them, so even objects in principle observable by the virtue of having been once inside the event horizon will eventually, and in a finite time, disappear from sight for all practical purposes.
(...)
* i.e. a distance you'd get if you froze the expansion of the universe and went out with a measuring stick to see how far something is - it's necessary to keep in mind what kind of distances we're talking about in cosmology, since there's a few 'types' (...).