Leaving the USS
If the USS is so bad, why don't employers and employees leave it?
They are pretty much trapped.
Once a defined benefit scheme is judged to be in deficit, any sane employer would want to run away from it, particularly if it a multi-employer scheme so someone else can pick up the bill.
An employer who stays will be forced to make extra contributions to pay off any alleged deficit, and will be forced to watch the scheme's assets being moved to lower-risk, but lower-yielding, assets, ensuring that the deficit is almost never paid off.
In the USS, the deficit contributions are a percentage of the pension contributions for current, contributing, employees. As such they are not directly related to the amount of the deficit which could be attributed to that employer's employees. An employer which can reduce its number of employees in the scheme will end up paying less towards the historic deficit, and vice versa.
To make things slightly less mad, two fundamental rules apply. An employer may not offer an alternative pension scheme to its USS-eligible employees. And an employer who does leave must pay a penalty charge based on an ultra-prudent (i.e. very exaggerated) estimate of the amount of the deficit which arises from his employees, both current and former.
At the time when Trinity College left, it had around twenty full-time employees in the USS (and a number of part-tiem employees and pensions which is unknown to me). Its exit bill was approximately £1m per full-time USS employee! The exit bill for Cambridge University is likely to be well over £1bn. It is safe to say that Cambridge will not be leaving on those punitive terms!
Post October 2021, for every £100 of gross salary that an employee gives to the basic USS scheme, the USS receives £318. That calculation is simple, but does the USS do anything useful with that money?
For contributions above the DB threshold, the calculation is simple. The USS puts £204 into one's DC pot, the other £114 going mostly into deficit reduction.
For contributions below the DB threshold, the calculation is hard. If it were easy, there would neither a deficit, nor arguments about how to calculate the deficit. The employee gets £13.60 pa of future pension, and £40.81 of future lump sum as the main benefits, assuming a 1/75 accrual rate. (For 1/85 the figures are £12 and £36.) For lifetime allowance tax calculations, this benefit has a value of £312 (£276), but that is a rather simplistic valuation, especially as it regards all indexation caps of at least 2.5% as being of equal value.
What is the DIY alternative? Under USS rules, the employer may not offer employer contributions to any other scheme for USS-eligible employees. It is probably also prohibited from offering salary sacrifice, so NI would still be payable on one's contributions, even though income tax would be refunded. So a standard rate NI payer would be able to invest £88 in a private pension in return for £100 of gross salary. (£98 for a higher rate payer, £86.75 and £96.75 from April 2022.)
Can a private scheme ever offset this loss? It depends on one's personal circumstances, including the length of time over which one can invest the money. Currently the core USS scheme has a single contribution rate of 9.8%, which some might find awkward to fund, whereas much lower rates are available in private schemes.
There are many other issues to consider too, such as death in service payments, ill-health retirement benefits, one's expected length of service in USS institutions, and whether one is likely to be UK resident for tax purposes in retirement. In particular, currently full USS benefits are obtained only after two years of contributions, and failure to reach two years service results in a sharp reduction in benefits. (This issue is due to be fixed in the April 2022 changes -- the one good thing about them!)
In summary, if you need to make a decision, there is no personal advice to be found on this wesite. If you want advice, speak to a professional advisor. It is not the case that the USS is clearly right (or wrong) for everyone.