On i386 docker image, pyperformance:

 ./python -m perf compare_to before.json after.json -G --min-speed=2
Slower (4):
- regex_dna: 279 ms +- 1 ms -> 288 ms +- 3 ms: 1.03x slower (+3%)
- deltablue: 17.9 ms +- 0.3 ms -> 18.3 ms +- 1.0 ms: 1.03x slower (+3%)
- scimark_monte_carlo: 258 ms +- 4 ms -> 264 ms +- 4 ms: 1.02x slower (+2%)
- unpack_sequence: 115 ns +- 1 ns -> 117 ns +- 3 ns: 1.02x slower (+2%)

Faster (9):
- scimark_lu: 408 ms +- 16 ms -> 389 ms +- 13 ms: 1.05x faster (-5%)
- spectral_norm: 315 ms +- 3 ms -> 302 ms +- 2 ms: 1.04x faster (-4%)
- xml_etree_process: 235 ms +- 2 ms -> 228 ms +- 3 ms: 1.03x faster (-3%)
- telco: 19.8 ms +- 0.4 ms -> 19.3 ms +- 0.3 ms: 1.03x faster (-3%)
- scimark_sor: 487 ms +- 6 ms -> 474 ms +- 4 ms: 1.03x faster (-3%)
- django_template: 385 ms +- 3 ms -> 376 ms +- 3 ms: 1.02x faster (-2%)
- unpickle_pure_python: 934 us +- 13 us -> 913 us +- 11 us: 1.02x faster (-2%)
- pathlib: 68.2 ms +- 0.8 ms -> 66.7 ms +- 0.5 ms: 1.02x faster (-2%)
- xml_etree_generate: 277 ms +- 3 ms -> 271 ms +- 7 ms: 1.02x faster (-2%)

Benchmark hidden because not significant (44)

and microbench:

 ./python -m perf compare_to before-dict.json after-dict.json -G    
Slower (2):
- fromkeys_str: 5.12 us +- 0.10 us -> 5.27 us +- 0.09 us: 1.03x slower (+3%)
- unicode_split: 636 ns +- 11 ns -> 648 ns +- 16 ns: 1.02x slower (+2%)

Faster (2):
- unicode_nodummy_5: 837 ns +- 28 ns -> 818 ns +- 21 ns: 1.02x faster (-2%)
- unicode_nodummy_6: 824 ns +- 10 ns -> 808 ns +- 20 ns: 1.02x faster (-2%)

Benchmark hidden because not significant (1): fromkeys_int