Not long ago i wrote this post about the future of database engines. It ringed a bell and reminded me a #IFTTD podcast by @Bruno Soulez podcast episod in which @Guillaume Rongier gave his own views for the future of databases and it was very close to what Stonrebraker and Pavlo concludes in their paper : databse engine will/should converge. He did also that the code execution would be reintegrated in the kernel of databases using shared memory to avoid the tcp stack. Mainly for edge computing but maybe not only.
At the very beginning databases were noSQL ( aka non relational ), then came the relational model and SQL with it and new needs finally brought back the noSQL on the forefront. Instead of going each their way, databse engine should converge