Not sure if "3-4 Sharpe" indicates the value of the Sharpe ratio you're earning since such magnitude is meaningless without some benchmark to compare against, due to it being a purely relative measure. Anyway, we can talk about 6 things that should stop you from using high leverage:
- Surprise bear market
- Asymmetric tail dependence
- Sharpe ratio misestimation
- In-sample performance exaggerates ($\neq$) actual performance
- Backtesting technique was wrong
- Prospect theory (asymmetric investment behavior)
It isn't hard to make positive returns, regardless of strategy, in a year that is providing 12% on the market. Are you confident enough to believe that your same strategy will still make 6% in a bearish economy that is providing -2% per year instead? Or will it triple the loss that the market is losing? Maybe you have thought that you don't stand to lose anything.
There are asymmetries to think about. More specifically, asymmetric tail dependence between asset returns, in that asset returns are likely to fall harder when the bearish regime comes around because of high correlation in the lower-tail-dependence than they were doing in the not-as-correlated upper-tail bullish regime which was the only scenario you backtested your strategy on. All you need to do is get the bull/bear bet wrong and the magnitude of risk-taking becomes less of an important indicator. Only the mere act of having borrowed any amount whatsoever in the first place becomes the regret in hindsight. Nevermind betting wrongly as to the direction of interest rate movements when they finally do turn non-zero. But you can look into bet-sizing if you want.
As a performance measure, the Sharpe ratio itself is flawed because it consists of asset means $\mu$, which is known to suffer from misestimation error. Setting up a strategy based on in-sample data, therefore, is likely to disappoint out-of-sample when statistics diverge from what they were in-sample. Minimizing variance alone, rather than the Sharpe ratio, is automatically more impressionable of a strategy since volatility $\sigma$ is easier to estimate than $\mu$, and therefore has superior out-of-sample performance than allocating based on the Sharpe ratio. If anything, the Sharpe ratio should only be used to evaluate performance after the fact, not to initiate trades.
Stating that you managed to construct a strategy that accomplished some feat should seriously be re-considered over in terms of how you properly you backtested it. Alot of backtests are done wrong leading to disappointment when a poorly-backtested strategy is actually implemented in real life.
As much as you might think that borrowing at 0% is free money, the reality is that you are likely going to be magnifying any losses you make. The more you have to put on the craps table, the more you are putting up to lose, your money or free money. Losing 10% of \$1,000,000 has a larger emotional impact than 10% of $10,000 no matter the source of funds, which is kind of explained by prospect theory in behavioral finance. Believing you are a guru in a bullish market is normal, so gains of any order are treated indifferently and happily accepted, unlike the two different loss scenarios just described, reflecting the asymmetry that the prospect of gains evokes in investment behavior versus loss aversion. Then again, maybe you have thought that you don't stand to lose anything.