The smart road is to continue all work (or at least work in progress) as a substantial number of mergers fail in the final stages. Even nTelos managed to put up a lot of LTE right before they were officially acquired. Hopefully it will be different this time for Sprint.
This merger could be very much like the Verizon-Vodaphone merger to setup Verizon Wireless. Softbank could end up with about 30% and Deutsche Telekom with 39%. Will Softbank acquire more shares later or quietly cash out?
Ultimately they will need to choose one of the EPCs and integrate around that. Then each enodeB will need to be pointed to the EPC and each integrated site will need to be assigned to each enodeB. VOLTE and SVRCC/WCDMA fallback will be the voice solution but will take time.