Doing some more analysis because I'm so obsessed with TQQQ. I have two portfolios in the PV simulations to backtest against OP's 55/45 allocation:
Backtesting dates: Jan 1987 to Jan 2019
Portfolio 1: 43/57 TQQQ-TMF (Max IR)
Portfolio 2: 31/69 TQQQ-TMF (Risk Parity)
Portfolio 3: 55/45 UPRO-TMF (OP)
In the first two scenarios, you can get the drawdown pretty close (in which most of the drawdown is coming from the dotcom bubble) to the financial crisis drawdown in Portfolio 3. Even the risk parity portfolio, which produces an exposure of 93% to QQQ dominates the 55/45 portfolio.
To address OP's & Moto's earlier concerns about the fund being wiped out in a single day, I don't really foresee that happening with the circuit breakers. Fund would automatically readjust leverage downward and be ok during the next trading day. I think with a maturing tech industry, one would want to include this in the allocation when it produces a higher Sharpe even including dotcom crash and using conservative approach on 3x leverage returns.
Also, I know we're probably beating a dead-horse when US Treasury stopped issuing callable bonds in 1985, but I really do think that skews the valuation of prior periods and have made the long end of the curve a decent attractive asset going forward. So much discussion has been dedicated to looking at 1955-1982, and I honestly can't see the point if treasuries were a completely different instrument during those time periods. Sorry if I missed the conclusion of that discussion