In the beginning of a run, gold will always be better. When you are just cruising along instakilling everything, extra DPS will not get you anywhere faster, but a gold bonus will allow you to level faster and instakill for longer. In the mid game, it kind of depends on the situation and the crusader. Panda/Robosanta and Kaine/Monkey are usually more advantageous than extra DPS just because their gold boost is so big if used properly.

My base gold bonus from all of my idols and trinkets is 264K%. In my current formation, I have a gold bonus of 7.01M%. That is about 25 times higher than base levels. It would take more than one or two levels to even out with that, and in the late game it usually takes more than one or two 15% DPS buffs to get even one extra level.

Testing to find your balance point is not all that hard to do. Set up a DPS heavy formation, get your main DPSer to, say, level 1000, and let it hit a wall. Let it sit at that level for a minute and check "gold gained in last minute" in the stats tab. Then change your weakest DPS buffer for your strongest Gold buffer, see if you get knocked back, wait a minute and check gold gained again. Repeat until satisfied. Takes 30 min tops.

When you're instakilling, neither gold nor DPS matter. The massive amount of gold you're making now will be insignificant 20 levels later, since prices increase geometrically.

I agree that while instakilling you should have every and any +%GP in group. Once you reach the point where your main DPS needs more GP than you collect in 4 areas to kill the boss however, you are left with a grind and it's not clear if +DPS or +%GP is strictly better.

I decided to waste some Sunday morning looking at it. The first issue is that GP does not scale in a easily understandable way. By charting the max GP per second (both by dividing amount earned by time during instakills and by calculating GP dropped per creature and average rate creatures spawn) I found the following pattern:

Area | % Increase from last area
1 | +0% (Same GPS as the last boss)
2 | +133% (or about x2.3 of the last area)
3 | +139%
4 | +56%
5 | +5%

Now, these numbers are all based on the first 25 stages and not that accurate but what is interesting is the difference in GPS between the same area in different zones. It's always close to x 9.25. In other 'words' the GPS of Zone 3 - Area 1 (Z1A1) = the GPS of Zone 2 Area 1 (Z2A1) * 9.25.

Not really that helpful if you need to grind to get GP but interesting to know.

Now, understanding the variation in GPS, you also need to understand the variation in DPS per area... this is even harder to measure and I certainly did a poor job but I'm only looking for ball park anyway as by the time you need these calculations you are dealing with a 4x difference in your DPS. A more accurate measure would probably be possible by finding the DPS needed to instakill for each stage. then also finding the DPS needed to instakill the boss...

Area | % Increase from last area
1 | -75% (The next stage is much easier than the boss)
2 | +150%
3 | +100%
4 | +100%
5 | +250%

This equates to the following assuming the last boss needed 100 units of DPS to kill and you earned 100 GPS during the boss area.
Area | DPS needed | GPS earned
0 | 100 | 100
1 | 25 | 100
2 | 62.5 | 233
3 | 125 | 557
4 | 250 | 869
5 | 875 | 912

Given this (probably flawed math) you should be able to clear area 4 with a single 25 level upgrade of your main DPS unit but will need AT LEAST 50 levels to clear the next boss (every 3 or 4 zones will need 75 levels or additional DPS contributions from other party members).

I suspect the numbers may change in higher levels (the curve may steepen) but don't have time to examine this anymore.

Latter if I have time, I'll test the numbers on latter levels then try to figure out what the trade off is between +%DPS and +%GP.