As others have come pretty close to saying, strongholds are in some respects a vestigial feature from early D&D. One major motivation for strongholds there was that eventually, players would get too powerful for any sane world to keep throwing challenging encounters at them, so they shift their focus to regional and global goals, toward which end holding territory is a logical component. It's part of the same mentality that said there could only be one 14th level druid per region (and later, one 15th level in the world). CRPGs have resolved this conundrum by creating insane hellscape worlds populated by implausible concentrations of high-level enemies with brain dead AI, so the player can defeat them, and we agree to ignore implications on the ecosystem. (More or less every CRPG that lets you get high-level does this; the later Might and Magic games had fun examples: "Wait, all these red goblins are kings? Where are their kingdoms? Did I just destablize goblin geopolitics by casting sparks?") BG2's Athkatla and its suburbs have two dragons, a full hive of beholders, and like seven high-level liches, all just waiting for you to bumble in so they can fight you to the death. All these enemies are supposed to be superhumanly smart. If there's a human GM controlling enemies that are supposed to be smart, this shouldn't regularly happen, and certainly not at sufficient frequency to let you regularly level up.
Similarly, the reason old D&D thought you needed to be involved at a political level to influence regional or global events is that those events tend to attract the attention of allied regional and global powers, and it's silly to think that their best response to a threat would be to send 4-6 bad dudes to punch their way through an army and then punch the enemy leader in the face. Again, CRPGs resolve this with the genre convention asserting that this is a totally smart thing to do. (If you want to focus on the political or global level, you can play a 4X game.)
So strongholds are in some ways solutions to problems that we nowadays just handwave over, and most games that have strongholds still rely on the CRPG standard method for dealing with major threats. Dragon Age 3 is a good example: you're commanding a paramilitary organization with multiple fiefs, you're allied with several heads of state, you have armies, squads of mages, and other powerful characters working for you, and the final sequence is still you and three of your bros running out to punch the final boss. Even in NWN2, everyone's example of a stronghold done well, what happens at the end? You successfully defend your stronghold against the evil armies . . . then four of you teleport into the final boss's lair and kill everyone in it. Your army, powerful allies, and indeed most of your party members sit around back home, braiding each other's hair and telling scary stories about the time they saw a goblin. ("He was red! Probably a king!")
From a narrative perspective, I like strongholds that ground my character's place in the world; giving the character some contested territory (MM7) or putting them in the path of the plot (NWN2) are good ways to do this. From a gameplay perspective, I like strongholds that are more like strongholds in strategy games with RPGlike elements, i.e. where building them out gives me some unique benefit that won't be eclipsed by mid-tier loot. PoE is an interesting example because pre-expansion, it got almost everything wrong, and a large number of these were fixed by the expansions: the location wasn't central to the action, investing in the stronghold didn't unlock any interesting bonuses beyond the bounty quests, the stronghold wasn't relevant to the main plot or indeed any subplots.