Why not just take on-element pillars out of the equation entirely and reduce the initial vault size to something like 360. If you want off-element pillars you'd still have to pack them.
Number of players would then be decided by dividing vault size by a smaller number, like 24 cards instead of 36. So if you had 48 cards, you'd have to field 2 players, but that isn't restrictive because they could each field 60 card decks loaded with pillars (since the limits don't include pillars).
Pillars are mostly an accounting headache as well as being a potential source of suicides, lack of pends also potentially limits salvaging creativity in an unnecessary way as well. They're also not all that important for opposing teams regarding strategy and vault tracking. What I cared about in vault tracking was specific cards or types of cards. I cared when an enemy ran out of pc, or cc, or shields or attacking creatures or something like that. Tracking pillars was largely unimportant.
So, any team can fill a deck with on-element pillars or pendulums anytime they want to, making necessary suicides non-existent. Instead it becomes all about the central/important cards, which is how it always was anyway, except now the pillar accounting problems are relieved.
Next, you eliminate pillar conversion entirely - since the on-element pillars are gone - and replace it with a new concept.
Change conversion to be something that happens when salvaging from a defeated opponent. Allow teams to either take the off-element card listed, or else convert it on the spot to on-element equivalent cards. An equivalent card would be along the lines xinef was talking about earlier, large quanta creature for large quanta creature, spell for spell, shield for shield etc. Essentially you categorize the types of card for conversion. This allows for a lot of on-element repair to vaults mid-war, and is quite powerful. As a result I think teams should have to publish their cards converted publically each round. That balances the additional power by revealing the choices for vault tracking by opponents.
This also makes vault tracking more important, because when a card is discovered to be in a vault, it doesn't "maybe disappear in conversion later", you know it's still in there.
The reason I think conversion of this type still has a place, is because you don't want the situation where a recovering team that gains an additional player to field through winning but has no on-element useful cards, and is simply loading decks with on-element pillars to qualify for 50% rules because that element will never be able to salvage on-element cards to fix it. That may not be a technical "suicide" but it still kind of is because it's so weakened by being composed mainly of half-dead pends.
The only additional issue this creates is you would need to have a minimum limit of non-pillar cards in each deck, say 12, and discards would be required to come from non-pillar cards as those are what are actually in the vault. You also couldn't discard more than was in the actual minimum limit in some decks because it couldn't be counted on to be there. Some types of decks (like stalls) are inherently pillar heavy, so this minimum limit couldn't be excessive or you'd break deck types. In order to keep the war from dragging on longer, you could still force 20-30 card discards (possibly 20 because pillars are now invisible so you don't need as high a discard limit). It's just that now 12 cards come from the deck used (the 12 is whatever your non-pillar minimum is), and the other 8-18 would be discarded from the vault. This would grant some leeway to teams using risky but powerful decks using key cards, but I'm not sure that's a bad thing. It would have the effect of making vaults harder to cripple through bad luck. Losing with the last deck containing deflags wouldn't necessarily mean you had to discard your last pc option, therefore your vault would be less easily crippled by untimely losses. That would lead to more potent fights for longer into the war as well.
Anyway, that's my 2 cents.
