Testing with low-memory conditions triggered fatal OOM crashes in the VM, as expected, but sometimes there were secondary crashes during error reporting with SIGSEGV, which suggests that memory exhaustion was not being checked somewhere. It would be preferable to not have these secondary crashes. E.g.

[error occurred during error reporting (printing all threads), id 0xc0000005, EXCEPTION_ACCESS_VIOLATION (0xc0000005) at pc=0x00007ffc107c99d0