Replaying trace takes a lot longer than recording

Summary

We're trying to use rr to debug a crash in Julia's garbage collector. However, replaying the trace takes too long, and makes debugging infeasible.

Commands

Recording the program with:

rr record --chaos julia -g2 mwe.jl

takes around 5-6min.

I've left rr replaying the program with:

rr replay

and it takes more than 15min.

Not sure if it matters, but I've also run this command in GDB:

handle SIGSEGV nostop noprint

to ignore segmentation faults than happen when Julia stops threads for GC.

Microarchitecture

Architecture:                x86_64
  CPU op-mode(s):            32-bit, 64-bit
  Address sizes:             46 bits physical, 48 bits virtual
  Byte Order:                Little Endian
CPU(s):                      36
  On-line CPU(s) list:       0-35
Vendor ID:                   GenuineIntel
  Model name:                Intel(R) Xeon(R) Platinum 8124M CPU @ 3.00GHz
    CPU family:              6
    Model:                   85
    Thread(s) per core:      2
    Core(s) per socket:      18
    Socket(s):               1
    Stepping:                4
    BogoMIPS:                5999.99

Reproducer

Unfortunately I can't share mwe.jl here due to licensing, but @Keno might have access to it.

Start with the reported rr record --chaos julia -g2 mwe.jl and rr replay commands, and obtain the unavailable mwe.jl reproducer if possible. Compare recording and replay timing while checking whether the GDB SIGSEGV handling affects behavior; done means identifying a reproducible cause or actionable fix for the replay slowdown.