HSE06 does give very good results for most systems, especially if you're interested in the band gap. However, it's at least an order of magnitude more expensive than GGA calculations, so I wouldn't recommend it for geometric optimization. You should be able to get quite accurate structures using GGA + U or even standard GGA, and then consider a more sophisticated method for a high-quality one-step calculation of electronic properties.

In general, if your system has more than 80 atoms I think it will be very hard to do even a one-step HSE06 calculation at a suitable level of accuracy (memory also starts to become a problem, at least in VASP). Interface structures are often larger than this, so you might need to use GGA + U instead, but this introduces the problem of choosing a good value of U.

Here are a couple of articles which look useful.

Article Density Functional Theory in Transition-Metal Chemistry: A S...

Article When Density Functional Approximations Meet Iron Oxides

I just have some additional comments:

- HSE06 belongs to the screened hybrid functional (XC) class of DFT.

Generally, it is cheaper than the "regular" hybrid XC ( for example PBE0,) due to the separation of the exchange into short (from the "exact" exchange") and long range (from PBE exchange functional) parts. It is expected to give similar performances on many properties (band gap, .

A reference for screened hybrid XC for solid-state chemistry and physics:

1. https://pubs.rsc.org/en/content/articlelanding/2009/cp/b812838c#!divAbstract

- For band gap, HSE06 is often considered as a very accurate XC. However, recently many studies have shown that all the hybrid XC perform very similar. So, the good prediction of band gap is apparently just due to adding the 'exact' exchange to the energy functional (can be seen as correcting self-energy interaction or over-delocalization of GGA functional or simply to find a good balance between under- and over-estimation of band gap behavior of DFT and HF, respectively).

Here are refs on this topic:

1. Article Predicting Band Gaps with Hybrid Density Functionals

2. Article Resolution of the Band Gap Problem for Materials Design

hope this is helpful

In my experience, HSE06 has given very accurate results of electronic properties in semiconductors like halide perovskites. However, since it is much more time-consuming and demanding in computer power, you can safely (and quickly) calculate structural properties with GGA and use HSE06 on top of them for electronic properties. HSE06 'heaviness' is not a problem if you can have access to multi-node HPC clusters your computational software is highly parallelizable.