As things stand right now with everything with Fortran across multiple processors, should a production codebase not explicitly include impliicit none statements then it is obviously risky to rely on a compiler-specific option only such as -fimplicit-none with gfortran. Given the implicit mapping scheme in the standard with I-N starting letters on object names rendering them default integer types and all else as default real can cause havoc to state the least.
A standard either means something or it doesn’t. I believe the way to address the nonsense with implicit none in Fortran is via the standard itself, not compiler options, nor fpm, neither LFortran, etc.
I do think Fortran 202Y standard revision, which practically can only come into effect at the earliest circa 2035, should consider making implicit none the default. One relatively simpler way to do so might be to eliminate implicit mapping once and for all.