My director offered to toss some offshore resources my way to help. Bluntly told him: I think I have 3 months work ahead of me. Give me offshore resources, it'll more likely be a 9-12 months, because by the time ive described the problem in sufficient detail, I could have just done the work. Instead, id have to review a broken "solution" that doesnt work, probably makes things worse.
Offshore is fine when you need grunt work, pounding out CRUD interfaces or already having a detailed spec. But, offshore doesnt work when you're flying by the seat of your pants.
Most of the big ossues I jad durong that release werent even my own. A few I remember: an x86/x64 inline ASM bug, use of an uninitialized variable in a DB lib that led to random failures dealing with NULLs, an overflow in a 3rd party datetime lib and ABI mismatch because of wrong datatypes used in an OSS ODBC wrapper that worked in 32-bit, but failed horribly on 64-bit.
None of the offshore talent I worked with at the time could have identified, let alone fix those issues. Several of those took reading the actual generated assembler to see what is going on. I dont write much x86/64, but I can read enough to know when shits going to hell.
The 2 big agencies I delt with were Salient and Accenture. I could work with Salient, most of their asserts were Indian and spoke English. Our Accenture resources were all Chinese. It was a pain in the ass because none of our engineering resources spoke or read/wrote English. Everything had to go both ways through a translator. Something was always lost in the translation. We ended up eating the contract and backing out with Accenture.