I'm not sure what would be the most useful way to bucket.
E.g. Fix the percentiles vs fix the size cut-offs. Perhaps a bit of both (fixed percentile, but don't allow cut-offs to be too close). 


...


Desired output:


Showing 531,506 symbols (402,348 unique) with total pss: 45003573 bytes
Stats for pss: 25% < 100, 50% < 500, 90% < 1000        <<<<<<< THIS LINE IS NEW
.text=35.2mb     .rodata=5.87mb     .data.rel.ro=1.82mb     .data=38.9kb     .bss=1.88mb     total=42.9mb
Number of unique paths: 16334

Section Legend: t=.text, r=.rodata, R=.data.rel.ro, d=.data, b=.bss
Index | Running Total | Section@Address | PSS | Path
------------------------------------------------------------
0)         47 (0.0%)  t@0x2f4e80   47 (size=188)  base/allocator/allocator_shim.cc
             operator new[] (num_aliases=4)
1)         94 (0.0%)  t@0x2f4e80   47 (size=188)  base/allocator/allocator_shim.cc
...


For a diff, it'd also be good to show number of increase vs decrease. e.g.:

1864 symbols added (+), 330 changed (~), 1612 removed (-), 519716 unchanged (not shown)
Of changed symbols, 112 grew, 218 shrank     <<<<<<< THIS LINE IS NEW
Number of unique symbols 402215 -> 402450 (+235)
11 paths added, 1 removed, 321 changed

Showing 3,806 symbols (aliases not grouped for diffs) with total pss: 20219 bytes
Stats for pss: 25% < -1000, 50% < -200, 70% < 0 90% < 200         <<<<<<< THIS LINE IS NEW
.text=17.7kb     .rodata=1712 bytes .data.rel.ro=256 bytes  .data=108 bytes  .bss=376 bytes  total=19.7kb
Number of unique paths: 339

Section Legend: t=.text, r=.rodata, R=.data.rel.ro, d=.data, b=.bss
Index | Running Total | Section@Address | Δ PSS (Δ size_without_padding) | Path
------------------------------------------------------------
- 0)      -3356 (-16.9%) t@0x0        -3356 (3352->0)    third_party/WebKit/Source/platform/graphics/GraphicsLayer.cpp
               blink::GraphicsLayer::LayerAsJSONInternal const