Intentionally and unintentionally leaving should be treated the same, as it's impossible to be sure which scenario happened in every case (e.g. a user plugging out their network cable versus the connection dropping). The only exception is to not wait for them if they were to explicitly select a "concede" option (but otherwise treat this option the same).
How much fun and how fast is ending a game (for each player)?
If ending the game is the most enjoyable part of winning (which it often is, for a lot of people), you want to allow them to continue until the end.
But an option to end the game (from either the losing players in the form of conceding, or the remaining players) may be useful if:
Ending the game is mostly a chore for both players.
A losing player is able to easily significantly delay the end of a game without really giving themselves a better chance of winning.
A winning player can indefinitely keep control of the game without ending it (which would be frustrating for the losing player).
How long are the games? How far along is this game?
Players may prefer finishing games if they've already committed a lot of time to the game and it's almost done.
They may prefer ending it if the disconnect happens after they barely started.
If the games only take a few minutes, you probably don't want to wait for disconnected players nor give remaining players a choice in terms of what happens (in order to keep the pace going). But either replacing them with a bot or just ending the game could make sense.
How many other players are there?
When there are multiple players in the game that needs to agree on something (whether it's the decision to concede or wait or continue with a bot), this could make things complicated and make players angry at other players or you. So one should be careful when implementing things like that.
How often do players leave? At which point in the game do they leave?
Why they leave could depend on a number of factors, including which device they're playing on, which country they're in, what penalties there might be for leaving and a number of other things specific to the game. But the bottom line is how often they end up leaving. You'll probably need to look at the data of your game, or at least your target audience, to determine this.
If the average player leaves often, anything other than silently replacing them with bots is likely going to be a huge irritation for the remaining players.
If players very rarely leave, you could make a good argument for waiting for them.
The effect on the game, and teammates
If there are just 2 players in the game, just ending the game could be a good option. But this is significantly worse when there are more than 2 players.
You really, REALLY want to avoid having a tie any time a player leaves or telling players they lost while the game is still ongoing (in case of team games or if you just pick the leading player as the winner).
If it's a team game, this would heavily favour a waiting option. Also, players may prefer to have their disconnected teammate just idle or die instead of being replaced by a bot which does "dumb" things.
Can players join a game as a team?
If players can form a team with friends, you'd almost certainly want to at least give the option to wait for disconnected players, and you probably don't want to replace them with a bot.
What's the penalty for losing versus simply leaving a game?
You may or may not want to treat these differently.
If simply leaving is heavily punished (either in terms of bans or just significant ranking changes), that would make a good case for giving a disconnected player some time to return to the game.
Is the game "casual" or "serious"?
In a "casual" game, it could make more sense to replace disconnected players with bots, whether silently or by asking the player first. You probably don't really want to do this with a game people are extremely passionate about, where they spend a lot of time trying to improve and there are tournaments and rankings and so on. For the latter, waiting for players would also make more sense.
What do your players think of playing with bots?
If players choose to play with other players, it stands to reason they prefer this above playing with bots (obviously).
How strong of a preference this is, however, and how this compares to other factors, would depend on the individual player, so this is something you could potentially A/B test.
How good are the bots?
This is especially important to ranked play, where players would have a certain expectation for the skill level of their opponents and you almost certainly don't want to replace players with bots, or at least this should be recorded as a win regardless of the actual result. But this may also be applicable to unranked play.
Having a bot that plays reasonably at all skill levels would likely be really difficult to achieve for a heavily skill-based game.
The bot should ideally be slightly worse than the player who left, and preferably play somewhat similar to how a human plays.
If they're better, this could suddenly make the game a lot harder, which will frustrate remaining players. If they're much worse, the game might become trivial to win, which may also frustrate players.
If they don't "act human", the switch between a human player and a bot may be rather jarring for the remaining players.
If they're either much worse or they don't "act human", silently replacing players with bots probably isn't a good idea.
What happens if players idle or "die"?
If a player can idle, die or have anything they control be removed from the game entirely without it making extreme changes to how the game is played for the remaining players, this might be a good option.
If the player has teammates, it might make sense to give any resources they might have or get to the other players instead, or allow their teammates to control their units.