We had the same issue in our past. For some reasons we managed that somehow. Because beside testing with simulators we also tested with real devices (since real devices have different behaviour than simulators).
First: Made an overview about the devices with the help of statistic webpage
In case of using real devices you have to consider that you are using:
- Different devices (Samsung, Apple, Huawei, Xiaomi, Motorola...)
- Mobile Versions (iOS 12.1, iOS 11.4, iOS 9.3, 8.0 Oreo, 7.0 Nougat...)
- Different usage of mobile and devices in different countries (e.g. percentage in Germany, Spain, France...)
Of course the business department wanted to test them all, but due to time circumstances it was not possible. Hence we focussed on the most relevant ones. We made a research via Statcounter:
All the datas are free via: GS Statcounter
Second: For overview purposes create browsermatrix and reduce the unrelevant ones:
Due to research in Statcounter we focussed on the most relevant ones and reduced the unnecessary devices/mobile versions.
In this case a browsermatrix like this helped in our case:

I know that this is just a small advice but in our case it helped to decrease the amount of testing devices. Of course there is a risk that some devices are not running / working. But in our case we couldn't test all combinations.
Third: Aske the business department for help before Prod with real devices
Some people are willing to help before Go-Live. Especially the business department was willing to help us since everybody has his private device ;-)