...perhaps we should embrace that aspect and make them much riskier.
It looks like Ighalli has done well to limit the number of god actions to four or five, which is good because it makes things easier to balance.
One of the features of the present system is that the Lawful gods tend to get you out of dicey situations, the neutral gods get you cool stuff and the chaotic gods pimp you with wicked powers to make you badass (but don't rely on them to save you). I like that a theme runs through the pantheon.
Holy books are all you need to change alignment, and there is no penalty for changing alignment other than forgoing benefits of other gods. Imagine if the way to increase relations would be via prayer only. Holy books would allow you only to learn about the god. The range of gods to be able to pray to would be say five, with highest chance of top relations with the centre god, and say 67% max relations with adjacent gods, 33% max with next adjacent gods. If you wanted to switch to a god with which you have negative relations, you would have to undergo a penance with that other god, so praying to them would mean you get punished until your relation with that god enters positive territory, and you start gaining better prayer effects. During this process you lose champion status and gradually that relation goes negative.
Once you reach champion, the adjacent gods lock in to 67%, 33%, 0% pattern, meaning adjacent gods give you the benefits of their lesser prayer outcomes in a tiered way.
Admittedly it sounds a bit like a frequent-flyer scheme.
It seems like there are different types of petitions (with relation requirement in brackets):
(0) common petition (pray for a new limb etc)
(1) god-specific effect 1 (0 - 33%)
(2) god-specific effect 2 (34 - 67%)
(3) god-specific effect 3 (68% - 99%)
(4) god-specific effect 4 (champion effect 100% relation)
You would be able to tell whether you can pray for an effect if it appears on the list of available petitions on the "prayer screen", giving you an idea of your current relation with that god.
You could forgo the champion effect and pray to two gods in exchange for their two tier-three effects, and on and on.
Positive relations could die away to zero (negative remain negative), so there is a reason to pray relatively frequently in order to keep relations up, until you reach champion.
One thing you could do is create two different types of behaviour, instant and delayed outcomes:
1) Instant outcome. You lose a limb or you are losing a battle. You pray, the god answers and saves your ass (or not, by some random factor).
2) Delayed outcome. There is no immediate danger, you pray to the god. He or she is busy, and they'll get back later.
2a) You get into shit, BAM! the god answers your prayer from before and saves your ass.
2b) After some random time (say within a bandwidth of 1000*(god-specific effect tier) - RAND_N(relation) or somethign) your prayer finally gets answered and you get what you prayed for (or maybe not, depending on some random factor). Prayer timer goes back to zero and then you know you can pray again for another effect.
If you pray multiple times (it resets the timer?) and you can change your petitition to the god, so the outcome of the prayer follows the last thing you prayed for. So high frequency prayer is pointless, but incurs no penalty. Means you can be a devout follower
(being punished for frequent prayer seemed counterintuitive).
The trick is, IVAN would need a good way of monitoring and prioritising what sort of help the player should receive, and whether it should be instant.
If you want to make the gods more capricious: So you want to be champion to a new god? Simply offer a limb at an altar, or renounce your old god for which you are currently champion (and face punshment by being scourged - scarred and covered in blood), or do battle with an armed warrior sent to kill you in `%gd`'s name. Or maybe you get sent a kamikaze dwarf to blow you up?
In fact, I think the type of kamikaze dwarfs should be restricted to the gods that you have negative relations with.
You can choose to pray whenever, no timer, but the amount of favour you have with the gods influences the result.
We could decouple the player's Lawful - Neutral - Chaotic meter to be restricted only to the types of deeds
that the player does, as opposed to being a status meter showing the player's overall alignment to a Lawful/Chaotic god + deeds. If the central, or champion god is say neutral, and the player's deeds tend toward Chaotic for example by eating human flesh, then when the player next prays, he may be hit with a punshment instead. So the player's behaviours need to align with their chosen god's morals. So if you're in with the chaos gods, you better be looking for evil deeds to do, just to stay in the gang. In this way, the player may be punished by prayer, but it is still within the player's control somehow, and gets to play a mini game to stay in control of their alignment.
It could also be that between when the player prays and when the prayer is answered, the difference in the values of the "Deed-alignment Meter" need to stay within some band in order for the prayer outcome to be positive. If the player has prayed to a lawful god when he was Lawful, but does evil and becomes neutral, then there is a negative gradient and the god may choose to punish the player.
Building a new god system is terribly difficult. The above is only one idea out of myriad different types of models you could implement. I hope it contributes to stimulating you own ideas. The game has extensive ancilliary machinery, like the danger system working in the background, which I haven't got a good handle on, but which can be utilised to assist the automation of some things. I think the main thing is that there is some structure nailed down in order to balance out the effects. Using a table or a matrix is usually a good way.