It's not the smartest move when you're the largest stakeholder. You or I as smaller stakeholders can grow our holdings by undermining the network at the expense of the larger stakeholders, and get a higher dollar value. The only way for the largest stakeholders to gain is to increase the value of the network. There's no getting around that, there's no way self voting becomes rational when you've got a major fraction of the total stake.
Although, in practice, even the smaller stakeholders have virtually all lost $ value in recent months, even with 100% self voting.
Hmm under linear, it's hard to trust other stakeholders not to vote-farm to outgrow their stakes disproportionately, so in the end, everyone does it taking the bet that the econs will be fixed at some point.
This is assuming that vote farming has 0 impact on the market value of the network. Which can be true temporarily, but is unlikely to be true for long. The market value of the network has far more impact on your net value as a large stakeholder than attempts to increase or maintain a % share.
The largest stakeholders don't actually have to trust that smaller stakeholders will not vote farm. They have the power to enforce it now, although I do agree that power could and should be made less costly (IMO the major costs are either human time or network resources, however Steemit Inc has not even really tried. @dan did in the past, and he tried to make it easier to enforce)
The point made by @smooth is not factored in to your analysis: concentrating SP rewards large accounts more than network effects diminish them presently.
An essential - and existential - problem is that stake-weighting forces all value other than financial to zero, both on Steemit and in the real world. It is necessary to measure and weight other values to effect rational society. Simply attending to finance alone creates a society of bots that have no other purpose.
People have other values, and many of those values are far more important than money.