Inspired by an unfortunate colleague, I am contemplating repairs and how to pay for them. One of the problems associated with our current project-centered approach to funding lab research is in repair costs. It is often not-cost effective (or practical) for an individual group to have the money to pay maintenance contracts. The really expensive equipment is often found in a user facility with user fees that theoretically cover maintenance, even if there are individual groups that would be capable of and willing to run the equipment alone due to this reason. Simple maintenance is actually not that terrible--most of the parts can be covered out the the budget for supplies. The problem comes in when there is an accident or major equipment failure in the lab. There are mechanisms for funding new equipment--wouldn't it be nice if some repair money came along with the instrument grant that was designed to be saved for a rainy day?
Repairs are CERTAINLY cheaper than buying a new machine (most of the time). However, in our project-based funding model, there is no way to really save up for a rainy day, when that rainy day is $15k or more in one shot with no advance notice. We had the same problem at National Lab, and mostly solved it in the the same ways people do at Universities--borrow from the department, borrow from other PIs who are unexpectedly flush, or go in the red (if possible), and pay it out of the supply budget from several projects in the next fiscal year.
At National Lab, it was actually fairly common to "carry over" money by helping out a fellow PI in the final year of your project, and having that PI pay for your salaries or supplies in a future year to pay back the debt (thus moving the money "forward" in time). This was especially common when a postdoc leaves in the last year of a project towards the end of the fiscal year, so there aren't many cost sinks in the ending project to soak up the extra cash (which gets pulled back if unspent in the final year).
Sometimes, I think it might be better to fund research programs and not projects. Then, it might be possible to have a lab savings account where extra cash can be saved for future repairs. That opens up a whole other can of worms (like how to avoid it being even less meritocratic/more "old boy" than the current system is a big one), but it might be better in terms of reducing waste in research spending, since less money would be spent on hoarding supplies that end up useless or on useless equipment just to spend down the budget instead of losing the money.
Research money is so tight these days, with labs unable to support existing personnel, let alone save something on the side for repair costs. It just seems so penny-wise and pound-foolish to keep ignoring this reality. It is in fact, something that keeps me up at night--that my really expensive workhorse instrument will break, effectively shutting my group down until I can scrape up the money to fix it.